Professional Documents
Culture Documents
A PALLAS PAPERBACK I 20
DEONTIC LOGIC:
INTRODUCTORY AND
SYSTEMATIC READINGS
Edited by
RISTO HILPINEN
~p~
~ paperbaCkS
Deontic logic.
In its modern form deontic logic, or the logic of normative concepts, has
been studied for about twenty years. The relevant literature consists
mainly of papers in various journals and publications, many of which
are not easily accessible to students of philosophy. Introductory text-
books have so far been missing. This volume has a two-fold purpose:
first, it brings together some important contributions to deontic logic,
and secondly, the papers published here have been selected in such a way
that they jointly can serve as an introduction to the problems and methods
of deontic logic. The essay~ by Fl1Illesdal and Hilpinen, Hintikka (partly),
and Segerberg were written specifically for this book; the others have been
published before. Some of these papers appear here in a revised form.
Most papers require some familiarity with elementary logic.
In the study of deontic logic, as well as other branches of what is now
commonly called 'philosophical logic', we can distinguish between two
main aspects. On one hand, deontic logicians have constructed formal
theories of various normative concepts, and on the other, they have tested
the adequacy of these theories by applying them to analysis of ethical
discussion. These applications may also involve criticism of ethical or
meta-ethical views. In this book, the former aspect is emphasized in
Kanger's paper, and the latter, e.g., in Hintikka's essay. A characteristic
feature of this book is the prevalence of semantical methods. These
methods are most likely to lead to significant further developments, not
only in deontic logic, but in philosophical logic and analysis in general.
All the papers included in this volume are by Scandinavian authors.
This book reflects the strength of interest in deontic logic and related
subjects in Scandinavia. In fact, Scandinavian philosophers have done a
great deal of the pioneering work in this field. The modern development
of deontic logic was initiated in the early 1950's by G. H. von Wright,
whose work stimulated most of the subsequent discussion of the subject.
Semantical theories of deontic notions were also presented first by
Scandinavian philosophers, Jaakko Hintikka and Stig Kanger (1957).
x PREFACE
THE EDITOR
INTRODUCTION TO
THE SECOND IMPRESSION
The articles included in this collection represent what may be called the
standard modal approach to deontic logic (the logic of normative concepts),
in which deontic logic is treated as a branch of modal logic, and the
normative concepts of obligation, permission (permissibility) and prohibi-
tion are regarded as analogous to the 'alethic' modalities necessity, possi-
bility and impossibility. In his recent paper [16] Simo Knuuttila has
shown that this approach can be traced back to late medieval philosophy.
Several 14th century philosophers observed the analogies between deontic
and alethic modalities and 4iscussed the deontic interpretations of various
laws of modal logic. A relatively simple deontic system of this kind (called
the system D or K D; cf. Lemmon and Scott [17], pp. 50-51, Chellas [10], p.
131) is obtained by adding to propositional logic two deontic axioms (or
axiom schemata),
(K) O(A :::J B) :::J (OA :::J OB)
and
(D) OA :::J ~ 0 ~ A,
where '0' is the obligation operator, and the deontic variant of the 'rule of
necessitation'
(0) From A, to infer ~A.
This system is closely related to the familiar alethic system T: the latter is
obtained from D by replacing '0' by its alethic counterpart '0' and
by strengthening the schema (D) into
(T) 0 A :::J A.
The system D is often called 'the standard system of deontic logic'. (The
standard system can be formulated in different ways; cf. pp. 13 and 127-128
of this volume.) More generally, any deontic system which includes the
system D may be termed a 'standard system'.
xii INTRODUCTION TO THE SECOND IMPRESSION
In the present volume, the papers by Stig Kanger and laakko Hintikka
outline the basic semantics of deontic statements, and laakko Hintikka
also shows how deontic logic can be brought to bear on various problems
and issues of moral philosophy, e.g. the is-ought question, the 'sollen-
konnen' principle, and the distinction between absolute obligations and
prima facie obligations. Georg Henrik von Wright (in 'A New System of
Deontic Logic'), Bengt Hansson, and Krister Segerberg discuss the pro-
blems of conditional obligation and commitment: in their contributions they
develop systems of dyadic deontic logic, in which the dependence of the
normative status of an act on the circumstances in which it is performed or
on the agent's earlier acts is expressed in terms of a dyadic obligation
operator (O-operator). In his second paper ('Deontit Logic and the Theory
of Conditions'), Georg Henrik von Wright suggests that deontic logic can
be construed as a fragment of the modal logic of (necessary and sufficient)
conditions, and argues that this way of looking at deontic logic leads to
illuminating analyses of the concepts of commitment and strong (disjunc-
tive) permission.
Much of the recent work on the logic of norms has centered on certain
'paradoxes', that is, arguments and examples which seem to conflict with
the basic principles of the standard approach. For example, Georg Henrik
von Wright's.and Bengt Hansson's work on conditional obligation -(re-
ported in the present volume) has been motivated by the 'paradoxes of
commitment' (see F~llesdal and Hilpinen's paper in this volume, pp. 23-24)
and by Roderick M. Chisholm's 'paradox of contrary-to-duty obligation'
(see pp. 24-25, 105 and 132-133). Chisholm's example is one of a group
involving conflicting obligations. The schema (D) or the principle of consis-
tency of the standard system excludes the possibility of genuine normative
conflicts; thus examples involving seemingly conflicting obligations pre-
sent a problem for the standard approach. In the case of Chisholm's
example, von Wright and Hansson solve the problem by relativising the
concept of obligation to circumstances: they assume that mutually incom-
patible obligations are relative to different conditions or circumstances. In
the 1970's the logic of conditional obligation has been studied by Bas C.
van Fraassen [25], David Lewis ([ 18], pp. 96-104, [19]), and by Azizah al-
Hibri, whose monograph [3] contains a survey of various deontic para-
doxes and a critical review of the theories of conditional obligation pro-
posed by von Wright, Hansson, van Fraassen, Lewis, and others. Brian F.
INTRODUCTION TO THE SECOND IMPRESSION X111
Chellas ([9] and [to], pp. 275-276) and Peter Mott [21] have presented
especially perspicuous analyses of conditional obligation in which the
ideas of obligation and conditionality are separated from each other, and
conditional obligations are defined in terms of the standard (monadic)
obligation operator and a dyadic conditional operator. Such an analysis
resembles that presented by Krister Seger berg in the present volume,
except that the conditional connective employed by Chellas and Mott is
not a strict or necessary conditional, but a variably strict conditional (in the
sense of David Lewis [18]; see p. 13), and consequently the concept of
conditional obligation defined by Chellas and Mott does not satisfy the
(objectionable) augmentation principle
principle (D) does not hold. The axiom schema (K) implies (by rule (0) and
propositional logic) the conjunction principle
(C) (OA & OB) ::::> O(A & B),
which cannot regarded as valid if the consistency principle is rejected:
perhaps a person can in some situation be subject to mutually incom-
patible obligations, but self-contradictory obligations are clearly impossi-
ble. (The standard system does not distinguish the consistency principle
from the principle that ought implies can, which denies the existence of
impossible obligations.) Neither (D) nor (C) belongs to what Brian F.
Chellas calls 'the minimal deontic logic' ([10], p. 202). Bas C. van Fraassen
[26] and P. K. Schotch and R. E. Jennings [22] have presented s~mantical
analyses of ought-statements which do not entail the validity of (D) or (C),
and which thus allow the possibility of genuine moral conflicts. The
problem of the resolution of nonnative conflicts has been investigated by
Carlos Alchourron and David Makinson [2], who show how a partial
ordering of a system of regulations may be used to resolve the inconsis-
tencies within the system.
In the standard approach to deontic logic, the concept permission
(permissibility) is defined simply as the absence of prohibition. The feature
ofthe standard system has been criticized on several grounds; for example,
it has been argued that the standard definition of permissibility excludes
the possibility of incomplete or open normative systems, that is, systems
which leave the normative status of some acts or states of affairs completely
undetermined. (Thus the standard approach cannot explain the distinction
between 'closed' and 'open' systems; cf. this volume, p. 166; von Wright
[27], Chapter IV; [28], pp. 413-415; and Carlos Alchourron and Eugenio
Bulygin [1], pp. 116-144, for discussions ofthis issue.) Perhaps the most
striking argument against the standard analysis of permissibility is pro-
vided by the 'paradox offree choice permission' or the 'paradox of disjunc-
tive permission' (see pp. 21 and 160 of this volume). Usually a disjunctive
permission seems to entail the permissibility of both disjuncts, but such an
inference is not justified by the standard system, in which 'PA' entails 'P(A
V B)" but not conversely. (Here 'P' is the permission operator.) Hans Kamp
[14] has presented an illuminating analysis of this problem in which he
shows that it is related to the peiformative use of permission sentences, that
is, the use of permission sentences for making previously prohibited acts or
states of affairs permissible and thus changing a normative system. In the
INTRODUCTION TO THE SECOND IMPRESSION xv
BIBLIOGRAPHY
[1] Alchourron Carlos E. and Bulygin, Eugenio, Normative Systems, Springer Verlag, Wien
and New York, 1971.
[2] Alchourron, Carlos E. and Makinson, David, 'Hierarchies of Regulations and their
Logic', New Studies in Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Publ. Co.,
Dordrecht, 1981.
[3] al-Hibri, Azizah, Deontic Logic: A Comprehensive Appraisal and a New Proposal,
University Press of America, Washington, D.C., 1978.
[4] Aqvist, Lennart, and Hoepelman, Jaap, 'Some Theorems about a "Tree" System of
Deontic Logic', in New Studies in Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Publ.
Co., Dordrecht, 1981.
[5] Castaneda, Hector-Neri, 'On the Semantics of the Ought-to-Do', in Semantics of Natural
Language, (ed. by Donald Davidson and Gilbert Harman), D. Reidel Publ. Co., Dor-
drecht, 1972, pp. 675-694.
[6] Castaneda, Hector-Neri, Thinking and Doing, D. Reidel Publ. Co., Dordrecht, 1975.
[7] Castaneda, Hector-Neri, 'The Paradoxes of Deontic Logic: The Simplest Solution to All
of Them in One Fell Swoop', in New Studies in Deontic Logic, (ed. by Risto Hilpinen), D.
Reidel Publ. Co., Dordrecht 1981.
[8] Chellas, Brian F., 'Imperatives', Theoria 37 (1971),114--129.
[9] Chellas, Brian F., 'Conditional Obligation', in Logical Theory and Semantic Analysis:
Essays Dedicated to Stig Kanger on His Fiftieth Birthday, (ed. by Soren Stenlund et al.),
D. Reidel Publ. Co., Dordrecht, 1974, pp. 23-34.
[10] Chellas, Brian F., Modal Logic: An Introduction, Cambridge University Press, Cam-
bridge, 1980.
[11] Greenspan, Patricia, 'Conditional Oughts and Hypothetical Imperatives', Journal of
Philosophy 12 (1975),259-276.
[12] Hilpinen, Risto, 'Deontic Logic and the Semantics of Possible Worlds', in Deontische
Logik und Semantik, (ed. by A. Conte, R. Hilpinen, and G. H. von Wright), Athenaion,
Wiesbaden, 1977, pp. 82-88.
[13] Hilpinen, Risto, 'Disjunctive Permissions and Conditionals with Disjunctive Antece-
dents', in Intensional Logic and Natural Language: Proceedings of the Second Soviet-
Finnish Logic Conference, Moscow 1979, (ed. by Ilkka Niiniluoto and Esa Saarinen),
Acta Philosophica F ennica, Helsinki, 1981.
[14] Kamp, Hans, 'Free Choice Permission', Aristotelian Society Proceedings N.S. 74 (1973-
74),57-74.
[15] Kamp, Hans, 'Semantics versus PragMatics', in Formal Semantics and Pragmatics for
Natural Languages, (ed. by F. Guenther and S. J. Schmidt), D. Reidel Publ. Co.,
Dordrecht 1979, pp. 255-287.
[16] Knuuttila, Simo, 'The Emergence of Deontic Logic in the Fourteenth Century', New
Studies in Deontic Logic, (ed. by Risto Hilpinen), D. Reidel Publ. Co., Dordrecht 1981.
[17] Lemmon, E. J. and Scott, Dana, The 'Lemmon Notes': An Introduction to Modal Logic,
(ed. by Krister Segerberg), American Philosophical Quarterly Monograph No 11, Basil
Blackwell, Oxford 1977.
[18] Lewis, David, Counterfactuals, Basil Blackwell, Oxford, 1973.
[19] Lewis, David, 'Semantic Analyses for Dyadic Deontic Logic', in Logical Theory and
Semantic Analysis: Essays Dedicated to Stig Kanger on His Fiftieth Birthday, D. Reidel
Publ. Co., Dordrecht 1974, pp. 1-14.
INTRODUCTION TO THE SECOND IMPRESSION xvii
The word 'deontic' is derived from the Greek word '3e6VtOl<;', which may
be translated 'as it should be' or 'duly'. Bentham uses 'deontology' for
"the science of morality", and Ernst Mally [30] was the first to use the
term - in the form Deontik - to refer to logical study of the normative use
of language. In accordance with Bolzano and Quine's definition of logical
truth, deontic logic can be defined as the study of those sentenc:es in
which only loaical words and normative expressions occur essentially.1
Normative expressions include the words 'obliption', 'duty', 'permission',
'right', and related expressions. These expressions may be termed Montic
words, and sentences involving them deontic sentences. a A deontic sen-
tence is a truth of deontic 1000c if it is true and remains true for all varia-
tions of its non-logical and non-deontic words (that is, expressions which
are not logical or deontic words). Deontic logic is closely related to the
logic of imperatives (or the logic of commands); in fact, many authors
regard these fields as essentially the same. 3 What is here called deontic
logic has also been referred to as logic of obligation and logic of norms
(or logic of normative systems).4
R. Hllpinell (ed.). Deolltic Logic: Introductory tmd SY8tellUltic Readlna8. 1-35. All right8 re8erved.
Copyright IC> 1970 by D. Reidel Publl8hillg ColfllHlllY. Dordrecht·Holll1llll.
2 DAGFINN FeLLESDAL AND RISTO HILPINEN
Axiom (A4') says that there exists a state of affairs u which ought to be
the case. Mally calls such a state of affairs 'the unconditionally obligatory'
or 'unconditionally required'. In Mally's system 'u' is, however, a propo-
sitional constant; thus, according to customary systems of quantifi-
cational logic where the quantifiers range over ordinary individuals,
DEONTIC LOGIC: AN INTRODUCTION 3
(10) states that, if something ought to be, whatever is the case ought to
be the case. But worse follows:
(12) states that whatever is the case, ought to be, and according to (20),
the converse implication is also valid. Theorem (21), the conjunction of
DEONTIC LOGIC: AN INTRODUCTION 5
(12) and (20), states that p ought to be the case if and only if it is the case.
(21) expresses the equivalence of ought and is (cf. [30], pp. 25 and 34).
These theorems are strongly counter-intuitive. Mally himself observes
that (21) is undoubtedly the strangest one among the 'strange' theorems
of his system ([30], p. 25).
He does not, however, reject his system of deontic logic on the basis
of theorems such as (12), (20) and (21). He constructs a philosophical
theory by means of which he endeavors to show that these consequences
are not as unnatural as they seem to be. For instance, Mally says that
his principles modify the presystematic notion of ought in important
respects; the presystematic notion involves ambiguities which are un-
covered by his logic ([30], p. 25). Nonetheless, it should be obvious that
results such as (21) are fatal to Mally's theory. In the first place, they are
in conflict with what Mally himself says about the notion of ought (or
obligation). For instance, Mally points out that
(22) Op v Oq
and
(23) 0 (p v q)
are not equivalent: (22) implies (23), but not conversely (cf. theorem (7»
([30], p. 27). This, of course, is what we should expect, but not so in
Mally's system. For in that system, (21) implies that (22) and (23) are
equivalent. A more serious consequence of (21) is that it makes deontic
logic trivial; according to (21), deontic logic is reducible to (non-modal)
propositional logic. By virtue of (21), cOp' is replaceable by 'p' every-
where it occurs; consequently Mally's axioms (Al)-(A3) reduce to tautol-
ogies of propositional logic, and both (A4) and (AS) are equivalent to u.
Where did Mally go wrong? It may be suggested that the inadequacies
of his system are attributable to a failure to draw a distinction between
(logical) implication and if-then statements (conditionals). For instance,
as was pointed out above, theorem (2) looks somewhat 'strange'. It is
an immediate consequence of (AI); hence axiom (AI) is not acceptable.
However, (AI) becomes much more plausible if the second conjunct of
the antecedent ('q:::l r') is replaced by 'r is logically implied by q' (or 'r is a
logical consequence of q'). Mally reads 'p :::I q' as 'p implies q'. The word
'implies' is notoriously ambiguous; in ordinary language it usually means
6 DAGFINN F0LLESDAL AND RISTO HILPINEN
Mally's system was criticized by Karl Menger in 'A Logic of the Doubtful.
On Optative and Imperative Logic' ([32], 1939). According to Menger,
Mally's main error was to attempt to build deontic logic upon the basis
of classical two-valued propositional logic. (This diagnosis has not been
borne out by later work in the field.) Menger suggests that deontic logic
should be based on three-valued logic, which includes, in addition to
the customary truth-values true andfalse, a third value, doubtful. Menger
constructs such a 'logic of the doubtful', and builds his imperative and
optative logic (the logic of wish) upon this three-valued system. Menger
has attempted to clarify the problems of ethics by formal methods also
in his book Moral, Wille und Weltgestaltung ([31]; 1934).
In addition to Menger's paper, there were published in 1939 papers on
imperative logic, a subject closely related to deontic logic, by Albert
Hofstadter and J. C. C. McKinsey [21], and by Rose Rand [39].8 Hof-
stadter and McKinsey's imperative logic can be criticized in the same
DEONTIC LOGIC: AN INTRODUCTION 7
way as Mally's deonticlogic: in their system cOp' and 'p' are equivalent (or,
in the terminology employed by the authors, !p and p are equivalent). In
'Zur Logik der Sollsatze' ([13]; 1939), Kurt Grelling presented a system
of deontic logic which includes the following axioms:
(B1) (p & q => r) => (p & Oq => Or)
(B2) (p => Oq) => (Op => Oq)
(B1) is essentially the same as Mally's axiom (AI), and leads to similar
unacceptable consequences. This was pointed out by Karl Reach in [40].
Reach observed that if we replace 'q' by '", p' and Or' by 'p', we obtain
from (BI)
(25) (p & '" p => p) => (p & 0 '" p => Op).
The antecedent of (25) is a logical truth; thus (BI) implies
(26) P & 0 '" p => 0 p ,
'p', 'q', etc. as referring to states of affairs; in von Wright'~ and Grelling's
systems they must be interpreted differently. Von Wright observes that
the word 'act' is ambiguous; it may be used to refer to "act-qualifying
properties", that is, general characteristics of acts (for instance, theft),
but also to individual acts, e.g., individual thefts ([46], pp. 1-2). Acts in
the former sense may be termed generic acts; acts in the latter sense are
termed act-individuals (Cf. [49], p. 36). In von Wright's system, deontic
operators are prefixed to names of generic acts; thus the symbols 'p',
'q', ... used above must be interpreted as standing for names of ge-
neric acts or act-predicates. This interpretation is philosophically sig-
nificant, and it also has certain purely syntactical consequences. If
deontic operators are prefixed to names of acts, ,the iteration of oper-
ators is not permissible: formulae such as 'Pp' and 'Op' are not act-
predicates, and consequently, e.g., 'OOp=.Op' and 'OPp :::l Pp' are not
well-formed formulae. For the same reason, 'mixed formulae', that is,
formulae in which propositional connectives are used to combine deontic
and non-deontic components (e.g., 'p :::l Op'), are not accepted as well-
formed. According to von Wright, it is, however, meaningful to speak
of the negation-act of a given act and of conjunction-, disjunction-, im-
plication-, and equivalence-acts of two acts. Thus, propositional logic is
applicable to, analysis of the logical relationships between generic acts.
Von Wright's system of deontic modalities is a decidable theory. The
validity of deontic formulae can be determined by a truth-table method.
In view of the restrictions imposed upon well-formedness of deontic
formulae, all well-formed formulae of von Wright's system have the
form (here we assume that all occurrences of '0' have been replaced by
'", P '" ')
where 'F' represents a truth-function, and fl' /2, ... , In are formulae
of propositional logic. Let Pl' P2' ... , p" be all sentential letters (i.e.,
atomic formulae) occurring in/l ,f2, ... ,fm. Let d; be the perfect disjunc-
tive normal form offj in terms of Pl' P2' ... , P., and let c~, c~, ... , ct
be
the conjunctive parts of d/. According to the principle of extensionality,
(27) is equivalent to
DEONTIC LOGIC: AN INTRODUCTION 11
(43) OOp::::> Op
DEONTIC LOGIC: AN INTRODUCTION 15
Recent studies by Stig Kanger [25], Saul A. Kripke ([271, [28]), laakko
Hintikka ([19], [20]), Richard Montague ([33], [34]), W. H. Hanson [15],
and others indicate that the application of modern semantical methods
(model theory) to deontic logic gives a more fruitful basis for understand-
ing deontic formulae and judging their acceptability. These semantical
theories concern quantified deontic logic, that is, deontic extensions of
first-order functional logic. Kanger's and Hintikka's theories are pre-
sented in detail elsewhere in the present volume; here we shall discuss
only the basic ideas of a semantics for propositional deontic logic. These
basic ideas are common to Hintikka's, Kanger's and Kripke's semantics.
The main difference between Hintikka's theory and the theories presented
by Kanger and Kripke is that the basic semantical device of Hintikka's
theory, a model set, is a set of formulae, whereas the models studied by
Kanger and Kripke are set-theoretical structures of a type more common
in modern semantics. The exposition below follows most closely Kripke
[27] and Hanson [15].
As was mentioned in Section III, deontic logic can be regarded as a
logical theory of normative systems. The principles of deontic logic
determine conditions of consistency for normative systems. By a 'norm-
ative system' we understand here simply any set of deontic sentences
closed under deduction.
When is a set of deontic sentences consistent? It seems natural to
require that at least the following 'minimal condition' should be satisfied:
(El) If a set of sentences A is consistent and {O 11> 012, ... ,0/.. , Pg}
s; A, then {fl' 12, ... , I .. , g} is consistent.
(El) says that a set of obligations is consistent only if all obligations in
this set can be simultaneously fulfilled, and that g is permitted only if it
can be realized without violating any of one's obligations. This seems
very plausible from the intuitive point of view. In fact, (El) (together
with rules for propositional connectives) is all we need for the standard
system of deontic logic: a deontic formula is provable in the standard
system if and only if its negation is inconsistent according to (El). It
should be observed that (El) does not require that all permitted states
of affairs can be realized simultaneously, but only that each permission
is compatible with all obligatory states of affairs. This shows that we are
here dealing with a fairly weak sense of 'permission': 'p is permitted'
DEONTIC LOGIC: AN INTRODUCTION 17
(F7) says that R is transitive. This condition has the same logical force
as the condition 20
(FS) says that if an obligation holds in our actual world, it also holds in
its deontic alternatives. If this condition is accepted, (42) can be shown
to be valid. Moreover, it may be argued that ifMl is a deontically perfect
world (or an ideal world), then it should satisfy, not only such obligations
as hold in M, but also those which hold in Ml itself. In other words,
(F9) If there is a world MeS such that R(M1' M), then R(M1' Ml)'
DEONTIC LOGIC: AN INTRODUCTION 19
Given this condition, (43) and (44) can be shown to be valid. 21 On the
other hand, it cannot be required that the relation R be reflexive in the
whole set S: R (M, M) is not an acceptable requirement, since our actual
world cannot be regarded as a deontically perfect world (on logical
grounds).
In [1] and [2], Alan Ross Anderson has suggested that deontic logic can
be reduced to alethic modal logic by means of a reduction schema
(G 1) 0 p == N (,..., p :::> S),
where 'S' is a propositional constant, and N is the modal necessity
operator. If it is assumed that the operator N satisfies, e.g., the axioms
of the Feys-von Wright modal system M,22 the axioms of the standard
system of deontic logic can be derived from (Gl) and the axiom
(G2) ,..., NS.
Anderson interprets'S' as a 'bad thing' or a sanction which results from
violation of one's duties. According to (Gl), p is obligatory if and only
if ,..., p (necessarily) implies the sanction S, in other words, p is forbidden
if and only if it implies the sanction. (G2) says that the sanction is avoid-
able, i.e., not everything is forbidden (or obligatory). According to (Gl),
the notion of permission (or permissibility) is defined by
(G3) Pp == M(p & ,..., S),
where M is the alethic possibility operator (Mp=="'" N,.., p). Thus, a state
of affairs p is permissible if and only if it is compatible with the absence
of the penalty S.
In an unpublished paper written in 1950, Stig Kanger presented a
simpler reduction schema,
(H1) Op == N(Q :::> p),
where 'Q' is a propositional constant, interpreted by Kanger as 'what
morality prescribes' (see [25], p. S3 in the present volume). According
to (HI), p is obligatory if and only if it is entailed by what morality pre-
scribes. Kanger's and Anderson's reduction schemata are closely related
20 DAGFINN F0LLESDAL AND RISTO HILPINEN
to each other: If '8' is equivalent to',.." Q', (Gl) and (HI) are equivalent.
Given (H 1), the axioms of the standard system of deontic logic are deriv-
able from principles of alethic modal logic and the axiom
(H2) MQ,
which says that morality cannot require impossible states of affairs. The
notion of permission can now be defined by
(H3) Pp == M(Q & p).
Thus, p is permissible if and only if it is compatible (or 'compossible')
with 'what morality prescribes', i.e., with all obligatory states of affairs.
It has been argued that the notion of permission defined by (G3) is too
weak. (G3) says only that it is possible to do the permitted thing and
escape punishment. "Must it not, however, be as certain that the man
who does the permitted is not punished for what he has done as it is that
he who neglects his duty is punished?" (von Wright [54], p. 90). This
criticism is beside the point, however: the 'sanction' 8 is not relative to
specific acts, but results from the performance of any forbidden act. If
a person performs a permitted act p, then surely he should not be punished
for p, but he may at the same time perform another act which is forbidden,
and thus implies 8. Performance of a permitted act cannot guarantee an
escape from 8, but it preserves the possibility of such an escape.
Anderson's reduction schema may also be criticized on the grounds
that violation of one's duties does not invariably lead to punishment (cf.
von Wright [54], p. 90). From a purely formal point of view (Gl) is
quite unobjectionable; the above criticism does not concern the schema
(Gl) as such, but rather the interpretation of '8' as a penalty. 8 cannot
be a 'naturalistic' description of some actual penalty. G. H. von Wright
has suggested that '8' can be interpreted as liability (as opposed to
immunity) to punishment ([54], p. 93). Ifliability to punishment is thought
of as being involved in any violation of obligations on purely conceptual
grounds, this interpretation is, indeed, attractive. It may also be assumed
that 8 means only that the obligations included in a given normative
ct>de are not fulfilled, i.e., that deontically perfect circumstances are not
realized (cf. Anderson [3], pp. 345-347, and [4], p. 111; Prior [36], p. 146).
This interpretation of 8 seems to come closest to Kanger's interpretation
of Q as 'what morality prescribes'. These interpretations of (Gl) and
DEONTIC LOGIC: AN INTRODUCTION 21
The standard system has been critized on the grounds that it includes
'paradoxical' theorems. For instance, many philosophers have felt that
there is something paradoxical in the formula
(46) Op =:l O(p v q),
which is a theorem of the standard system. This theorem says that if a
certain state of affairs p ought to be the case, then also p v q ought to be
the case. For example, if I ought to mail a letter, I also ought to mail or
burn it. But if I in fact ought to mail a letter, then surely it is awkward to
say that I ought to mail or burn it. This example has been presented by
Alf Ross [44], and the paradox involved in it (and other similar examples)
is called Ross's paradox. A similar paradox can be formulated in terms
of the notion of permission as follows:
(47) Pp ::> P(p V q)
is a theorem of the standard system. According to (47), if a person is
permitted to smoke, he is also permitted to smoke or kill. Now a per-
mission to smoke may sound innocuous enough, but its alleged conse-
quence, a permission to smoke or kill, seems decidedly immoral.
These applications of (46) and (47) may indeed sound paradoxical,
but they lose most of their paradoxical character as soon as we pay
22 DAGFINN F0LLESDAL AND RISTO HILPINEN
The 'paradoxical' theorem (47) is not valid for this notion of permission.
According to von Wright, the notion of free choice permission cannot
be formalized in the standard system. It seems to us, however, that a free
choice permission can be expressed in the standard system in a perfectly
adequate way: Pp&Pq. If 'a is permitted to smoke or kill' is a free choice
permission, it should be formalized as Pp &Pq, and this is not, of course,
implied by Pp (according to the standard system). If the word 'or' is
interpreted in this way (as it often is in ordinary language), 'a is permitted
to smoke' does not imply 'a is permitted to smoke or kill'. There is no
need to invent special notions of permission or construct special logics
of permission and obligation on the basis of this accidental interchange-
ability of the words 'or' and 'and' in ordinary language. 27
All paradoxes discovered in the standard system of deontic logic are
not, however, as uninteresting as those discussed above. There is a group
of paradoxes which, so it seems to us, show something significant about
the limitations of the standard system. This group has variously been
called the paradoxes of derived obligation, paradoxes of commitment, or
paradoxes of contrarY-fo-duty imperatives. 28
As was mentioned above, von Wright [46] formalizes the notion of
commitment by formulae of the type
which has the same form as (46). Nevertheless, (50) suggests that (49) is
- at least in some cases - an inadequate formalization of commitment.
24 DAGFINN F0LLESDAL AND RISTO HILPINEN
In Formal Logic ([37], pp. 224-225), Prior suggested that the notion of
commitment may be formalized by formulae of the type
This alternative was not available to von Wright in his 1951 system, since
(52) is not a well-formed formula in that system. However, if deontic
operators are prefixed to sentences, (52) may be accepted as well-formed.
If the notion of commitment is formalized as (52), the paradox corre-
sponding to (50) does not arise;
(57) says that if not p, then the man ought not to do q, that is,
(61) ,...., p=>O ,...., q;
and (58) is simply
(62) ,...., p
Chisholm points out that (59-62) imply a contradiction. According to the
standard system, (55) and (56) imply
(63) Oq,
whereas (57) and (58) imply (by Modus Ponens)
(64) 0 ,...., q.
I
notion of permission, P(p r), which may be read as 'p is permissible
under circumstances r'. The system has two axioms:
(66) ! !
P (p v q r) == P (p r) v P (q r). I
I
According to (11) and (66), under constant circumstances pep r) and
I
O(p r) satisfy the axioms of von Wright's original (1951) monadic
system.
In [50] von Wright presented another system of dyadic deontic mod-
alities, which includes the following axioms:
(K2) I
,.., (O(p r) & 0(,.., pi r»)
(K3) 0 (p & q ! r) == 0 (p ! r) & 0 (q I r)
(K4) O(p I r v s) == O(p I r) & O(p ! s)
In the same way as in the case of the monadic system, we may exclude
'empty' normative systems by adding to (K2)-(K4) the axiom
(K5) I
P (p r) == '" 0 ('" p I r).
(Kl )-(K3) say that under constant circumstances the notion of condi-
tional obligation satisfies the principles of the monadic standard system.
Axiom (K4) is more interesting; it is specific to conditional obligations.
In [51] von Wright mentions that the construction of this system was
stimulated by Chisholm's discussion of contrary-to-duty imperatives
(pp. 103-104; p. 115 in this volume). Below, we shall test the adequacy
of von Wright's system by applying it to Chisholm's example. 31
28 DAGFINN F0LLESDAL AND RISTO HILPINEN
worlds are a subset of all deontically perfect worlds. Thus the following
modification of (72) is valid:
(76) O(p It) & P(r It) ::> O(p r).I
More generally, it can be shown that according to (Ll), (~4.2) must be
weakened to
(77) I
O(p r v s) & P(r s) I ::> I
O(p r).
Bengt Hansson has recently proposed for conditional deontic modalities
a semantical foundation of the type sketched above ([16]; see pp. 143-
146 in this volume). Hansson's semantical system combines the logic of
preference with deontic logic. This theory is in accord with Chisholm's
requirement: according to Hansson's theory, it is possible to give reason-
able answers to the question of what we ought to do after we have failed
to fulfill our 'absolute' obligations. It may be expected that semantical
theories of this type will eventually provide a solid intuitive foundation
for systems of conditional deontic modalities.
University of Oslo
University of Turku
BIBLlOGRAPHy33
[1] Anderson, Alan Ross, The Formal Analysis of Normati"e Systems (Technical
Report No.2, Contract No. SAR/Nonr-609 (16), Office of Naval Research,
Group Psychology Branch), New Haven 1956. Reprinted in The Logic of Decision
and Action (ed. by N. Rescher), University of Pittsburgh Press, Pittsburgh, 1967,
pp. 147-213.
[2] Anderson, Alan Ross, 'A Reduction of Deontic Logic to A1ethic Modal Logic',
Mind 67 (1958) 100-103.
[3] Anderson, Alan Ross, 'Some Nasty Problems in the Formal Logic of Ethics',
Nous 1 (1967) 345-360.
[4] Anderson, Alan Ross, 'Comments on von Wright's "Logic and Ontology of
Norms"', in Philosophical Logic (ed. by J. W. Davis, D. J. Hocmey, and W. K.
Wilson), D. Reidel Publ. Co., Dordrecht, 1969, pp. 108-113.
[5] Bar-Hillel, Yehoshua, 'Imperative Inference', Analysis 16 (1966) 79-82.
[6] Broad, C. D., 'Imperatives, Categorical &. Hypothetical', T71ePhilosopher 1 (1950)
62-75.
[7] Castafteda, Hector-Neri, 'The Logic of Obligation', Pliilosophical Studies 10 (1959)
17-23.
[8] Chisholm, Roderick M., 'Contrary-to-duty Imperatives and Deontic Logic',
Analysis 24 (1963) 33-36.
32 DAGFINN F0LLESDAL AND RISTO HILPINEN
[9] Dawson, E. E., 'A Model for Deontic Logic', Anolysis l' (1959) 73-78.
[10] Feys, Robert, Modal Logics (ed. with some complements by J. Dopp), Gauthier-
Villars, Paris, 1965.
[11] Fisher, Mark, 'A Logical Theory of Commanding', Logique et Analyse 4 (1961)
154-169.
[12] Gombay, Andre, 'Imperative Inference and Disjunction', Analysis 2S (1965)
58-62.
[13] Grelling, Kurt, 'Zur Logik der Sollsiitze', Unity of Science Forum, January 1939,
44-47.
[14] Grice, H. P., 'The Causal Theory of Perception', Proceedings of the Aristotelian
Society, Suppl. Vol. 3S (1961); reprinted in Perceiving, Sensing and Knowing (ed.
by R. Swartz), Doubleday and Co., New York, 1965, pp. 438-472.
[15] Hanson, W. H., 'Semantics for Deontic Logic', Logique et Analyse 8 (1965)
177-190.
[16] Hansson, Bengt, 'An Analysis of Some Deontic Logics', Nous 4 (1970) 373-398.
Reprinted in this volume, pp. 121-147.
[17] Hilpinen, Risto, Rules of Acceptance and Inductive Logic (Acta Philosophica
Fennica 22), North-Holland Publ. Co., Amsterdam, 1968.
[18] Hilpinen, Risto, 'An Analysis of Relativised Modalities', in Philosophical Logic
(ed. by J. W. Davis, D. J. Hockney and W. K. Wilson), D. Reidel Publ. Co.,
Dordrecht, 1969, pp. 181-193.
[19] Hintikka, Jaakko, Quantifiers in Deontic Logic (Societas Scientiarum Fennica,
Commentationes Humanarum Litterarum 23:4), Helsinki 1957.
[20] Hintikka, Jaakko, 'Deontic Logic and Its Philosophical Morals', in Models for
Modalities. Selected Essays (by J. Hintikka), D. Reidel Publ. Co., Dordrecht,
1970, pp. 184-214.
[21] Hofstadter, Albert and McKinsey, J. C. C., 'On the Logic of Imperatives',
Philosophy of Science 6 (1939) 446-457.
[22] Hohfeld, Wesley Newcomb, Fundamental Legal Conceptions as Applied in Judical
Reasoning (ed. by W. W. Cook), Yale University Press, New Haven, 1919.
[23] Hughes, G. E. and Cresswell, M., An Introduction to Modal Logic, Methuen and
Co., London 1968.
[24] Jl!lrgensen, Jl!lrgen, 'Imperatives and Logic', Erkenntnis 7 (1937-8) 288-296.
[25] Kanger, Stig, New Foundations for Ethical Theory, Stockholm 1957. Reprinted in
this volume, pp. 36--58.
[26] Kanger, Stig and Kanger, Helle, 'Rights and Parliamentarism', Theoria 32 (1966)
85-115.
[27] Kripke, Saul Aaron, 'Semantical Analysis of Modal Logic I: Normal Modal
Propositional Calculi', Zeitschrift fur mathematische Logik und Grundlagen der
Mathematik 9 (1963) 67-96.
[28] Kripke, Saul Aaron, 'Semantical Considerations on Modal Logic', Proceedings
of a Colloquium on Modal and Many-Valued Logics (Acta Philosophica Fennica
16), Helsinki 1963, pp. 83-94.
[29] Levi, Isaac, Gambling with Truth, Alfred A. Knopf, New York, 1967.
[30] Mally, Ernst, Grundgesetze des Sol/ens. Elemente der Logik des Willens. Leuschner
& Lubensky, Graz, 1926.
[31] Menger, Karl, Moral, Wille und Weltgestaltung. Grundlegung der Logik der Sitten,
Wien 1934.
[32] Menger, Karl, 'A Logic of the Doubtful. On Optative and Imperative Logic',
DEONTIC LOGIC: AN INTRODUCTION 33
NOTES
• Sections I-V of this paper are written by Risto Hilpinen on the basis of lecture-notes
by Dagfinn Ffllllesdal; sections VI-IX are written independently by Risto HiJpinen.
This work has been supported by grants from the Finnish Cultural Foundation
(Suomen Kulttuurirahasto) and the Finnish National Research Council for Social
Sciences (Valtion yhteiskuntatieteellinen toimikunta).
1 Cf. Quine [38], p. 387.
2 This expression has been used by C. D. Broad in [6].
3 See Fisher [11], Kanger [25], sections 2.3 and 2.4 (pp. 44-50 in this volume), von
Wright [53], p. 154, Aqvist [55], p. 248, and [57], p. 183 n. 1.
4 Cf. Anderson [1] and Castaneda [7].
5 Our notation here differs from Mally's; our u corresponds to Mally's U, n to n,
w to V, and m to /\.
6 In [30], this rule is applied e.g. on p. 29.
7 This is not the only possible explanation. In [19] Iaakko Hintikka has pointed out
that if the material implication q ::::> r in the antecedent of (AI) is replaced by a deontic
implication O(q ::::> r), we get an acceptable principle of deontic logic. Cf. also the
formula (41) below.
8 These papers, as well as the paper by GrelIing mentioned below, are reviewed by
Frederic B. Fitch in The Journal 0/ Symbolic Logic 5 (1940) 39-42.
9 Von Wright's 1951 system of deontic logic is described briefly also in [47], chapter 5.
10 These principles are theorems in all normal systems of alethic modal logic. For
various systems of modal logic, see Feys [10] and Hughes and Cresswell [23].
11 Cf. Iaakko Hintikka [19], p. 16, [20], and pp. 59-104 in this volume.
12 Cf. Thomas Aquinas, De Veritate, Q 17, art. 4, and Summa Theologiae, IaIIae,
Q 19, art. 6, and IlIa, Q 64.
13 We owe this term to Mr. Bengt Hansson; see [16) (p. 122 in this volume).
14 Cf. von Wright [46], pp. 1-2.
15 Cf. Iaakko Hintikka [19], pp. 7-9.
16 See e.g. [52], p. 16.
17 If this interpretation is accepted, the expression 'to do p' is, of course, somewhat
deviant. In [52) G. H. von Wright has suggested that it can be replaced by 'to see to it
that p' (p. 16). For simplicity, we shall use also the former expression regardless of the
interpretation of sentential letters.
18 cr. Hintikka [20] and pp. 79-82 in this volume.
1~ This weak notion of permission may also be termed 'permissibility' (cf. Hintikka
[20]). It must not be confused with the notion of right or claim as defined by Hohfeld
in [22], pp. 36--38. If a person a has against another person b a right (or a claim) that p,
then b has toward a a duty that p. Such obligations are not involved in the present
notion of permission. For a discussion of Hohfeld's analysis of basic legal notions, see
DEONTIC LOGIC: AN INTRODUCTION 35
Kanger's paper in this volume, and Kanger and Kanger [26]. Another formalization
of Hohfeld's system can be obtained by relativising deontic operators to persons in
the same way as alethic modal operators are relativised in [IS]. If a's duty toward b
that p is expressed by '0 ... ~p', a's right (or claim) against b that p can be expressed by
'O~ ...P'. In the systems of deontic logic discussed in the present paper, deonticmodalities
are not relativised to persons. This involves the assumption that we are speaking only
of the obligations and permissions of one single person (or a group of persons).
20 (F2)-(F4) and (FS) correspond to the following condition of consistency: If a set
of sentences A is consistent and {O/I, Ols, ... , 0/,., Pg} £ A, then {Oil, Dis, ... , 01.. ,
/1, Is, ... , I,., g} is consistent.
21 The theory of deontic modalities presented by Hintikka in [20] corresponds to that
defined by the conditions (F1)-{F4), (FS), and (F9).
22 For the system M, see [10], p. 124.
23 There has been a great deal of discussion of the proper interpretation of the operator
Nin (GI). Cf. Anderson [3], pp. 348-354, [4], p. 111, and von Wright [53], p. 147.
24 In Levi's theory of induction and rational belief, 'the strongest sentence accepted
via induction' corresponds to Kanger's constant Q (cf. [29], pp. 29-30, and [17],
especially p. 84). In Hilpincn [17], the sentence h* corresponds to Q. This analogy
results from the similarity of deontic logic and the logic of rational belief (or 'accept-
ability'); the logic of rational acceptance sketched in Section 4.1. of [17] is similar to
the standard system of deontic logic.
25 In [9] E. E. Dawson has presented a reduction in which Olis defined as MNland
PI as NM/, and in [56] Lennart Aqvist has discussed other reduction schemata of the
same type. These 'reductions' do not appear very interesting from a philosophical
point of view.
26 For a discussion related to this point, see e.g. Grice [14], Section m.
27 cr. Anderson [4], pp. l08-U)9. Similar confusions seem to underlie the recent
discussion of 'choice-oft'ering' and 'alternative-presenting' disjunctive commands by
Rescher and Robinson [43], Gombay [12], and Aqvist [57]. For a criticism of this
discussion, see Bar-Hillel [5].
28 Cf. Prior [35], and [37], pp. 224-225; Chisholm [S], and von Wright [49].
29 Cf. Hansson [16] (pp. 132-133 in this volume).
30 cr. Hansson [16].
31 In [41] and [42] Nicholas Reacher has presented another system of conditional
deontic modalities. This system is discussed and criticized in detail by Bengt Hansson
[16] (reprinted in this volume); we shall not discuss it here.
32 Dr Audrey McKinney has pointed out in correspondence (in 1973) that according to the
truth-dt:finition (LJ) given below, 'O(rjr)' is always true, and hence
O(p& -pjp& -p)
is a logical truth. Thus (K2) should be weakened into
(K2*) 0 r => -(O(pjr)& O( - pjr»,
where 'Or' means that r is true in at least one possible world.
33 This bibliography includes only those works mentioned in this paper. A comprehensive
bibliography of deontic logic is to be found e.g. in von Wright [52].
STIG KANGER
INTRODUCTION
There are philosophers who believe they really tell us how we ought
to behave, and there are philosophers who believe they ought to tell us
how we really behave. There are also philosophers of a more convenient
sort, who only wish to tell us something about ethical theories. The
doctrines advocated by these three kinds of philosophers all belong to
the wide domain of ethics, and they are easily confused. However, we
shall try not to confuse them here, and for that purpose, we may put
them under three distinct headings:
1. ethical theory proper
2. the psychology and sociology of morals
3. moral philosophy.
An ethical theory (of the proper kind) sets forth a system of ethical
propositions as true. Some ethical theories also set forth a system of
imperatives as correct. These propositions and imperatives are supposed
to give us the moral norms we are assumed to need.
We are all acquainted with an example of an ethical theory, namely,
the ten commandments.
An ethical theory may be more or less stringent. The stringency may
vary in several respects. For instance:
1. The propositions (and imperatives) set forth in the theory are all
formulated in a language with a certain formal structure. This structure
may be more or less specified.
2. The statements (and the imperative sentences) of this language are
either given or assumed to have a certain interpretation. This interpre-
tation may be more or less specified, and it may grant the propositions
(and the imperatives) a greater or less degree of lucidity.
3. Some statements (and imperative sentences) in the language are logi-
cally valid, and some are logical consequences of others. The methods for
demonstrating validity and consequence may be more or less developed.
R. Hllpinen (ed.). Deontic Logic: Introductory and Systematic Readings. 36-58. A.1I right. reserved.
Copyright © 1970 by D. Reidel Publishing Company. Dordrecht·Holland.
NEW FOUNDATIONS FOR ETHICAL THEORY 37
1. A Formal Language
In this section, I shall outline a formal language fit for the part of
ethical theory which we call the pure theory of norms. This language,
which we call L, is obtained from the formal language of lower predicate
logic by the addition of a modal operator 'Ought' and an imperative
operator 'I'.
The symbols of L are the following:
(1) Parentheses and commas
(2) Propositional constants (to be introduced when needed)
38 STIG KANGER
Deontic logic, i.e. a logic with some means for expressing the notion of
ought, originates with E. Mally, Grundgesetze des Sollens (Graz, 1926).
In this work Mally added a modal operator, denoting ought, to the
formal language of the propositional logic and laid down some (not very
happily chosen) axioms for this operator. Mally used the term UDeontik"
for what we call deontic logic.
The modalities Ought, Right and Wrong may be distinguished from
the predicates of being Obligatory, Permitted and Forbidden: the latter
apply to actions. A logical system involving these predicates has been
given in G. H. von Wright, 'Deontic Logic' (Mind 60 (1951) 1-15). The
statements of this system are built up by means of symbols for truth
functions and statements of the form PX expressing that some action of
the kind X is permitted (in short: X is permitted). The class of kinds of
actions is closed under the Boolean operations. Thus ',.., P X' and
, ,.., P ,.., X' may be interpreted as expressions of the facts that X is for-
40 STIG KANGER
planned joint human efforts. Thus it might well be the case, for instance,
that it Ought to be so that Mr. X saves the drowning man in front of
him, even if Mr. X is unable to do so because he does not know how to
swim.
My explanation of Ought expresses a kind of utilitarianism (which
roughly corresponds to my own vague ethical beliefs). But we shall not
emphasize this point too much. We could very well interpret welfare in
such a way that the welfare programs would satisfy, for example, the
extreme deontologist, and our explanation would still serve its purpose.
So much for Ought. We have defined the modality Right as '" Ought",.
Thus defined it is related in a certain way to Ought. But, if we conceive
Right as an explication of the vague notion of rightness, this relation is
by no means a matter of course. There are other relations which seem to
be equally natural. We may say, to give a single example, that something
is right if it is not Wrong, and if it involves something that Ought to be
the case but that is not triVially obtained. To be more explicit: We say
that the proposition It is right that A is true in the universe of discourse
if and only if (i) the proposition RightA is true in this universe and
(ii) there is a proposition P such that the three propositions: it is un-
avoidable that if A then P, OughtP, and it is avoidable that P are true
in the universe of discourse. We note that rightness in this sense - we
may denote it by 'Right+' - admits indifferent facts that are neither
Right+ nor Wrong.
The difference between Ought and Right on one hand and Right + on
the other may be illustrated by an example. Consider the following three
propositions:
Al : Thomas sends five pounds to Elisabeth, if he owes five
pounds to Elisabeth.
A2 : Thomas sends a five-pound note to Elisabeth, if he owes five
pounds to Elisabeth.
A3: Thomas sends a five-pound note to Elisabeth without owing
her anything.
The three propositions OUghtAl and Right+ Al and RightA l are true
(we may argue). Right+ A2 and RightA 2 are true, but OUghtA2 is false,
since Thomas could equally well have sent five pound notes instead.
RightA3 is true but Right+ A3 and OughtA3 are false.
42 STIG KANGER
We shall not confuse the notion It is right that with the notion of a
Right which we may have in mind when making propositions like:
Elisabeth has a right to get back the five pounds she loaned to Thomas.
Propositions of this kind may be regarded as idiomatic instances of the
schema
(0) X has a right in relation to Y to the effect that F (X, V).
Here X and Yare what we might call 'moral personalities' (to coin an
analogy to 'legal personalities'). If we consider various instances of this
schema we shall find that the prefix 'X has a right in relation to Y to the
effect that' is ambiguous. The following four idiomatic instances of (0)
suffice to illustrate this fact:
(1) X has a right to get back the money she loaned to Y.
(2) X has a right to walk into V's shop when it is open.
(3) X has a right to give all her money to Y.
(4) X has a right to walk on the street outside V's shop.
In (1) 'right' means claim, in (2) 'right' means liberty or privilege, in (3)
'right' means power and in (4) 'right' means immunity. Corresponding to
these four senses of 'right' are four alternative explications of the ambig-
uous schema (0):
(1') Ought (Y sees to it that F(X, V»~
(2') Right "'" (X sees to it that ,.., F (X, V»~
(3') Right (X see~ to it that F (X, V»~
(4') Ought"'" (Y sees to it that"'" F(X, V»~
Finally, we shall not confuse Ought and Wrong with the notions of
praiseworthiness and blameworthiness. We usually apply these latter
notions to people (or to moral personalities). We may say, for instance,
that I am blameworthy for writing a paper on moral philosophy. This
means, I think, at least six things: (i) It is Wrong that I write a paper on
moral philosophy, (ii) I am writing a paper on moral philosophy, (iii) I
can avoid writing a paper on moral philosophy, (iv) I can know that it
is Wrong that I write a paper on moral philosophy, (v) I can know that
I am writing a paper on moral philosophy and (vi) I can know that I can
avoid writing a paper on moral philosophy.
We readily see that the notion of Right +, and the different senses of
a Right, as well as the notions of praiseworthiness and blameworthiness
NEW FOUND A TIONS FOR ETHICAL THEORY 43
In ethical literature the notions of ought, right and wrong are usually
conceived as predicates applying to actions - sometimes they are defined
only for actions which are free in some sense. A wrong action is usually
conceived as an action that ought not to be performed. But there is less
agreement as to the relation between right and ought. Two examples
- one on the Right line and one on the Right + line - might illustrate this.
G. E. Moore defines ought, right and wrong only for actions that are
voluntary in the sense that they could have been avoided, if we had chosen
to do so. A wrong voluntary action is one that ought not to be performed,
and a right voluntary actio~ is one that is not wrong. We note that there
are no indifferent voluntary actions. See G. E. Moore, Ethics (London
1912), Chapter I; cf. also Principia Ethica (Cambridge 1903), p. 148.
W. D. Ross suggests that a wlong action is one that ought not to be
performed, and a right action is one that is neither wrong nor indifferent,
in the sense that it does not fulfill any moral claim. (See W. D. Ross,
Foundations of Ethics, Oxford 1939, Chapter III.)
The distinction between claim, immunity, power and liberty, as differ-
ent senses of a right, is well known in jurisprudence. The distinction has
been particularly elaborated by W. N. Hohfeld. (See W. N. Hohfeld,
Fundamental Legal Conceptions, New Haven 1919, Chapters I-II.) The
Hohfeldian scheme may be carried over to the domain of ethics and it
may be outlined in a table with four groups, each containing four
equivalent schemata. (Instead of Hohfeld's term 'no-right' we use
'exposure', following a suggestion made in J. R. Commons, Legal Foun-
dations of Capitalism (Madison 1924). We also use the term 'claim' instead
of 'right'.)
3. Valuation
I now turn to semantics. By a range, we shall understand a non-empty
class of individuals. We shall use the variable 'r' to denote ranges.
By a primary valuation (for L), we understand any binary operation V
(with the class of ranges as the first argument domain and the class of
constants and variables as the second) which, given a range r, assigns:
6. Normative Logic
Let the symbol 'M' denote anyone-place modality. Assume that
T(r, V, MA) is defined for each r and V and each statement A of L. Let
.£I be any empty or non-empty, finite, denumerable or non-denumerable
sequence of propositions of L. Let MLI be the sequence obtained from .£I by
prefixing 'M' to every statement contained in the proposition of .£I. Thus,
if .£I is (BI' VI), (B2' V2), then MLI is (MBl> VI), (MB2' V2).
Now the following two conditions are always equivalent:
(1) For each .£I and each (A, V), if .£I entails (A, V), then MLI
entails (MA, V).
(2) There is a relation R such that for each r and each (A, V),
T(r, V, MA)= 1 == (r') (R(r', r) ~ T(r', V, A)= 1).
The relation R is unique:
If for each r and (A, V)
T(r, V, MA) = 1 == (r') (R(r', r) ~ T(r', V, A) = 1),
then for each r' and r
R(r', r) == (B)(V)(T(r, V, MB) = 1 :::> T(r', V, B) = 1).
If M is Ought and if R is Roupt> then the first result provides a justifi-
cation of the valuation for the deontic statements. The second result
provides a kind of explanation of Roulh,. Perhaps we may paraphrase it
as follows: Roullht is the relation which holds between any two universes
r' and r such that every proposition that ought to be true in r is true in r'.
We shall now make three assumptions concerning the relation Roulht:
I (r)(Er') ROullhl(r', r)
II (Er) "'" ROup,(r, r)
III (Er')( (Er) ROulhl (r', r) & (Er) "'" ROupt (r', r»)
Assumption I is equivalent with the assumption that
(OughtA:::> RightA)
is always valid. Assumption III is equivalent with the assumption that
there are synthetic deontic propositions of the form (OughtA, V). It is
also clear that III is an expression of a kind of moral relativism.
NEW FOUNDATIONS FOR ETHICAL THEORY S3
I shall now list a few valid sentences. The validity of these sentences
does not depend on the assumptions just made.
Ought(OughtA :::> A)
7. A Dialogue
The complete moral philosopher: Excuse me for interrupting you, Mr.
Kanger, but would you admit a short interview before you proceed?
Kanger: Yes.
Ph: According to a wellknown theory in moral philosophy, known as
NEW FOUNDATIONS FOR ETHICAL THEORY 55
the emotive theory, deontic propositions are neither true nor false. Now,
I understand, you have the opposite view.
K.: Yes.
Ph.: Of course you and the adherents of this theory may have in mind
two different notions of truth. But your notion of truth seems to be in
agreement with scientific semantics and I am sure that the adherents of
the emotive theory would adopt it if they were met with the problem.
K.: Yes.
Ph.: Now, clearly, deontic propositions with the valuation they got in
Section 3 must be either true or false. So, if the adherents of the emotive
theory wish to sustain their standpoint they have to reject the valuation
clause for Ought or the equivalent thesis that OughtLi always entails
(OughtA., V) when LI entails (A., V), which you gave in Section 6.
K.: Yes.
Ph.: The equivalent thesis can, of course, be refuted by the argument
that deontic propositions are neither true nor false.
K.: Yes.
Ph.: I realize that this would be a petitio principii. But perhaps there
are other arguments for rejecting the valuation clause or its equivalent.
Let me try the argument that deontic propositions of the form
(Ought A. , V) do not state anything about reality. But this argument
involves, of course, nothing that might refute the valuation clause or the
idea that deontic propositions are true or false. At most, we may conclude
that deontic propositions are not synthetic.
K.: Yes.
Ph.: The feeling we may have that (Ou~htA., V) does not state any-
thing about reality is perhaps easily explained: Because of the vagueness
of Ought we may feel that (OughtA., V) is not synthetic, even if it is so,
and hence, we may also feel that it does not state anything about reality.
K.: Yes.
Ph.: Then I know of no other argument against your standpoint except
the one that deontic propositions are synonymous with imperatives. But
I also know your answer: Every deontic proposition is synonymous with
an imperative, and this fact is in full agreement with everything else in
this paper.
K.: Yes.
Ph.: Now let me return to the notion of truth. Perhaps our agreement
S6 STIG KANGER
on this point was a little rash. Could we not restrict the application of the
truth predicate to, say, non-deontic statements and call the formerly true
deontic statements correct instead? And couldn't we do this and still be
in agreement with scientific semantics?
K.: Yes.
Ph.: Do you mean that the choice of the range of applicability of the
truth predicate is, to some degree at least, conventional?
K.: Yes.
Ph.: So the adherents of the emotive theory have a chance to be right
after all by a terminological trick.
K.: Yes.
Ph.: I now turn to a new problem. There is a .wellknown distinction
between so called natural properties and non-natural properties. Some
authorities believe that value is a non-natural property, while others
disagree. In this paper, the distinction is difficult to maintain because of
your tendency to do away with all kinds of spurious entities. The predi-
cates do not refer to properties in the sense we may have in mind in this
connection, but to classes of individuals or to classes of ordered sets of
individuals. And I cannot see how one class can be less natural than an-
other. This fact does not, of course, mean that the distinction between
naturalism aJ?d non-naturalism cannot be maintained at all. We may
perfectly well distinguish between naturalistic and non-naturalistic state-
ments and propositions. Thus, we say that A is naturalistic if A or ,.., A
is logically equivalent with a sequence of ordinary statements, and non-
naturalistic otherwise. We say that <A, V*) is naturalistic if <A, V*) or
<,.., A, V*) is synonymous with a sequence of ordinary propositions of
<L, (fJ), and non-naturalistic otherwise.
K.: Yes.
Ph.: It follows from assumptions I and II of the preceding section that
there are non-naturalistic deontic statements. But we may raise the
problem whether or not there are non-naturalistic deontic propositions
of <L, (fJ). The answer to this problem depends on the choice of (fJ.
However, I think we need not hesitate to assume that there are non-
naturalistic deontic propositions of <L, (fJ). In any event, if we turn to
standard English and raise the analogous problem, then a negative
answer would certainly be unwarranted - it would be, we may argue, an
expression of the so called naturalistic fallacy. But the naturalist position
NEW FOUNDATIONS FOR ETHICAL THEORY 57
University of Uppsa/a
NOTE
• An earlier version of this essay was published as a privately distributed pamphlet:
New Foundations for Ethical Theory. Part 1, Stockholm 1957.
JAAKKO HINTIKKA
Pp = it is permissible that p.
R. HliplM1I (ed.), Deo1ltlc Logic: I11t.odIu:to.y tIIId SystemAtic Readings, 59-104. All rights .ese,.ed.
Copy.ight C> 1970 by D. Rdtkl Pooblishi1lg COmp01ly, Dordrecht-Holltllld.
60 JAAKKO HINTIKKA
In this paper we shall disregard the (highly relevant) fact tltat obligations,
and even more so duties, are usually thought of as obliging some particular
person, and that permissions often are similarly 'personal'. In this chapter,
and in its immediate sequel, the reader will have to think of obligations
in completely 'impersonal' terms. Most of our discussion will nevertheless
remain valid if it is assumed that all the obligations, permissions, prohi-
bitions, etc. we are considering relate to one and the same person. Whether,
and if so how, 'personal' obligations and permissions can be defined in
terms of the 'impersonal' ones is an interesting question which will not
be discussed here.
As bound (bindable) individual variables we shall use the lower-case
letters x, y, .... As quantifiers we shall employ the usual symbols '(3 x)'
and '(y)' (where x and y may of course be replaced by other bindable
individual variables). As place-holders for expressions referring to
individual acts (as 'names of act-individuals') we shall use the lower-case
letters a, b, c, ....
2. It is implicit in what was just said that the individuals we shall in the
first place consider are individual acts. It would be easy to introduce
variables and quantifiers ranging over some other kinds of individuals,
for instance Qver persons. By using such variables and quantifiers, im-
portant further observations can be made and codified. It seems to us
that these observations are nevertheless not quite as fundamental as
those that can be made in terms of variables ranging over act-individuals
(and in terms of place-holders for expressions which refer to such in-
dividual acts).
But must we really - and may we - use variables of the latter kind, that
is to say, deal with particular acts as being (logically speaking) individuals
over which we can quantify? Here I cannot do full justice to these ques-
tions. Fortunately, it so happens that Donald Davidson has argued my
case for me to a considerable extent in his paper 'The Logical Form of
Action Sentences' (in The Logic of Decision and Action, ed. by N. Rescher,
University of Pittsburgh Press, Pittsburgh, 1967, pp. 81-95). There he
argues incisively for the affirmative answer both to the question of
possibility and to the question of necessity of considering particular acts
as (logical) individuals. Suffice it here to refer the reader to Davidson's
arguments. It may be added, though, that insofar as we can elucidate
SOME MAIN PROBLEMS OF DEONTIC LOGIC 61
In the rule (F.3), the variables 'a' and 'x' may be replaced by other
variables of the same kind (free and bound, respectively). In the same way,
rules which are formulated in terms of particular variables are in the
sequel understood to apply to all variables of the same kind.
The rules (F.I}-{F.3) define the formulae we are going to use for our
purposes. In discussing the systems of other authors, however, we shall
frequently quote formulae of different kiIlds, too. For instance, in the
above rules the deontic operators '0', 'P', and 'F' are applied to formulae
and not directly to predicates of acts, while in other systems such a direct
application is allowed. The merits of our procedure are shown by the
arguments by means of which we shall try to show (in Sections 5 and 6
below) the indispensability of quantifiers in deontic logic.
3. As the reader may gather from the introductory survey paper of this
volume, most expositions of deontic logic fail to utilize quantifiers, i.e.
the precise logical counterparts of such words as 'every' and 'some'.
Sometimes quantifiers are brought into treatments of deontic logic as
additions to systems already constructed. Even when quantifiers are
62 JAAKKO HINTIKKA
(1) -0- p == Pp
(2) -P- p == Op
(3) -Op == P-p
(4) -Pp == O-p
(5) Op == F-p
(6) Fp == O""p
Roughly speaking, (1) may be read 'it is permissible thatp if and only
if it is not obligatory to omit p', (2) may be read 'p is obligatory if and
only if it is not permissible that not-p', and correspondingly for the other
types of formulae. In virtue of (1), 'P' may in the current systems be
replaced everywhere by''''' 0 -', and vice versa; and similarly for (2)-(6).
In virtue of (6), 'F' can always be eliminated in terms '0' and' - '.
I do not want to criticize (1)-(6) or the related possibilities of elimi-
nating '0', and 'F' and 'P' in terms of one another. In fact, I shall
assume the same interdependencies in my own approach. What I would
like to point out are the awkward consequences of replacing 'p' in (1)-(6)
by a variable ranging over properties of acts. In other words, I maintain
that one cannot subscribe to (1)-(6) while employing formulae like OA,
SOME MAIN PROBLEMS OF DEONTIC LOGIC 63
tween them, the other cannot be expressed within the system. Under the
interpretation (ii), many logical laws of the propositional calculus fail.
For instance, from a and '" a v b we can no longer infer b, from a and b
we cannot infer (a&b); etc.
Under the interpretation (i), these laws of the propositional calculus
are applicable. However, this interpretation has other undesirable conse-
quences. Under it, the meaning of the 'propositional' connectives '",',
'&', and 'v' is not the same within the scope of deontic operators and
outside of them. From this it follows that a formula of the form Pp may
mean something entirely different from saying that the state of affairs
expressed by p is permitted. For instance, under (i) '" a means that no
act of the kind A is done. Under the usual interpretation of deontic
operators, on the other hand, P", a does not mean that it is permissible
not to do any act of the kind A. It merely means that one is permitted
to do an act of the kind A, i.e. to omit A. Similarly, a &b means that an
act of the kind A is done and (independently of this) an act of the kind B
is done, while P(a&b) ordinarily means that one is permitted to do an
act with both the properties A and B.
The only way of obtaining a uniform interpretation of all our formulae
under the interpretation (i) is to change the meaning of formulae like
P(a&b) to say that one is permitted to do A and (independently of this)
to do B. Under this reading, formulae like P(a&b) == (Pa&Pb) should
be provable. They are not provable, however, in any current system. The
reinterpretation hence runs contrary to the intentions of the deontic logi-
cians. And for good reasons, indeed: the new interpretation would greatly
impoverish the applicability of deontic logic. Among other things, the
old interpretation is not expressible within a system in which the new
one is presupposed.
Thus, in order to obtain a satisfactory interpretation of Prior's for-
malism, we must choose the remaining alternative. We have to take
formulae like a&b to be concerned with some (arbitrarily chosen) partic-
ular situation. In other words, a&b should be read: 'An act of the sort A
is done in the particular situation we are considering, and an act of the
sort B is done in the same situation'. Under this interpretation alone is
one entitled to infer from 'An act of the sort A is done' and 'An act of
the sort B is done' that an act of both the sort A and the sort B is done.
This was seen to be indispensable if we want to obtain a uniform inter-
SOME MAIN PROBLEMS OF DEONTIC LOGIC 67
we are not speaking of this world only. Rather, we are speaking in some-
thing like a counterfactual or sUbjective mode. We are speaking of what
could be or could happen: we are saying that something could be done or
could be the case without violating any obligations (norms). In other
words, in saying that p is permissible in a world we are therefore saying
that this world can consistently be thought of as being replaced by another
in whichp is the case but in which all obligations are nevertheless fulfilled.
This implies that we cannot formulate conditions on the presence of
Pp in a description Jl of a possible world in the sole terms of Jl alone. In
addition to Jl, we have to consider another description Jl* of a possible
world, related to Jl in a certain way. This way will be expressed by saying
that Jl* is a deontic alternative to Jl. Intuitively, we may think of Jl* as a
description of that state of affairs in which p was assumed to be the case
for the purpose of showing that it can be the case while all obligations
are fulfilled. (Sometimes we shall also speak of the possible world de-
scribed by Jl* as an alternative to the world described by Jl.)
These observations motivate the adoption of the following condition:
(C.P*) If PpE Jl, then for at least one deontic alternative Jl* to Jl we
must have p E Jl*.
It -is, of course, assumed that deontic alternatives to a model set are
also model sets, i.e. that they satisfy the same conditions (C. '" )-(c. U)
as Jl.
This condition suffices to make sure that p can be the case, i.e. that there
is no inconsistency in assuming that p. But we wanted to accomplish
much more: we wanted to require that p can be the case while all obli-
gations are fulfilled. If this is to be accomplished by means of the deontic
alternative Jl*, we have to impose further conditions on it. We have to
require that whatever ought to be the case in the world described by Jl is
the case in the world specified by Jl*. In other words, we have to adopt the
following condition:
(C.O+) If Op E Jl and if Jl* is a deontic alternative to Jl, then p E Jl*.
(C.o*) If Op E JJ., then for at least one deontic alternative JJ.* to JJ. we
must have p E p.*.
10. The main complication we have to take care of here is that the set JJ.*
mentioned in our conditions may in turn contain formulae of the forms
Pp and Op. In order to be sure that JJ.* is satisfiable we must hence apply
our conditions to p.* and not only to JJ.. The rule (C.P*) then compels us
to consider a third set p.** of formulae which is a deontic alternative to
JJ.* but not necessarily to JJ.. Again we may find it necessary to apply the
same condition to JJ.** so as to necessitate the consideration of still more
sets of formulae, etc.
For these reasons, we must define the notion of satisfiability in terms
of a set Q of sets instead of just two sets. Combining all the above obser-
vations, we may say that a set A of formulae is satisfiable if and only if it
can be imbedded in a member of a model system. A set Q of sets is a model
system if and only if it satisfies the following conditions: (I) every member
of Q satisfies the conditions (C.-)-(C.U); (II) a dyadic (two-place)
relation (calle.d the relation of deontic alternativeness) may be defined
among the members of Q in such a way that (i) each member JJ. of Q
satisfies (C.P*) and (C.o*) with respect to some member JJ.* of Q: (ii) any
two members JJ., JJ.* of Q satisfy (C.O+) and (C.OO+); (iii) each member
JJ.* of Q satisfies (C.O)rest if JJ. is in Q.
Further conditions (over and above those used to define a model
system) do not seem to be forthcoming. For instance, permissions ('per-
missibilities') cannot in any case be 'moved over' from JJ. to one of its
alternatives, say A, in the way (C.OO+) says that norms are 'movable',
for obviously one can quash a permission by making use of another.
It is important to realize that, and why, we often have to consider more
than one deontic alternative to a given world, say M. This is due to the
fact that not all permissions that obtain in M can be made use of in the
same world. The simplest case in point is a deontically neutral proposition
p, i.e. one for which PpEP., P - pEJJ.. Since p and - p cannot both be the
case in the same possible world, (C.P*) forces us to consider more than
one deontic alternative to JJ..
SOME MAIN PROBLEMS OF DEONTIC LOGIC 73
and those that would obtain in such an alternative POSt ible world, are
assumed to be fulfilled in each of them.
Notice also that for Kant the categorical imperative is obviously the
principle of all maxims, both of those obtaining in the actual world and
of those that are followed in a 'Kingdom of Ends'. Thus the requirement
that all these maxims are followed presumably implies that in a Kantian
'deontically perfect world' both the 'old' and the 'new' obligations are
fulfilled, just as our conditions (C.O·) and (C.O)rest require.
Generally speaking, the deontic alternatives to a given world are re-
lated to it rather in the same ways as a Kantian 'Kingdom of Ends' is
related to the actual world. From the point of view of this given world,
they are realizations of the (normative) ideals obtaining in it somewhat
in the same way as a Reich der Zwecke. We can say of them the same as
was said by Kant of a notion closely related to that of a Kingdom of
Ends, namely of the notion of an intelligible world (Verstandeswelt):
"The concept of an intelligible world is therefore only a point of view
(Standpunkt) which the reason finds necessary to adopt in order to think
of itself as an active being" (urn sich selbst als praktisch zu denken ... op.
cit., p. 458, Kant's italics). We could perhaps bring Kant's formulation
a little closer to ours if we changed the last few words to read: in order
to be able to think of itself as acting according to its normative principles.
Once Kant says that in morality a possible world of ends is considered as
if it were the actual world of nature (op. cit., p. 436, footnote).
In a sense the notion of a deontic alternative therefore is thus a some-
what watered-down and relativized variant of the Kantian notion of a
'Kingdom of Ends'. It is a much weaker notion because our concept does
not contain any reference to a particular moral principle, be it the cate-
gorical imperative, universalizability, or what not, in the way Kant's
notion does. It is a relativized notion, for it refers to the possible world
which a deontic alternative is alternative to (from the point of view of
which it is so to speak considered). Moreover, a deontic alternative to a
given possible world is not a unique entity, contrary to the way in which
Kant seems to have looked upon his 'Kingdom of Ends'. On the contrary,
normally there are several deontic alternatives to a given possible world
in a model system. This difference between us and Kant reflects partly
the more important role which is played by the notion of permission in
our thinking as compared with Kant. For it is precisely the multiplicity
SOME MAIN PROBLEMS OF DEONTIC LOGIC 75
(A. E) If Il e Q' and (3 x) p E Il but not p (b/x) E Il f<'r any free vari-
able 'b', then we may adjoinp(a/x) to Il, provided that 'a' is
a free variable not occurring in the formulae of any member
of Q'.
(A.U) If Il E Q' and (x)p E Il but not p(b/x) e Il for some free vari-
able 'b' occurring in the formulae of Il, then we may adjoin
p (b/x) to Il.
(A.P*) If Il E Q' and Pp E Il but not p E Il* for any member Il* of Q'
which is a provisional alternative to Il, then we may adjoin
{p} to Q'.
We shall say that the new member {p} of Q' is a provisional (deontic)
alternative to Il. This relation of provisional alternativeness is supposed
not to be disturbed by adjunctions by the rules (A) to Il or to {p}. We
also assume that all the relations of provisional alternativeness obtaining
between the members of Q' are originally due to an application of the
rule (A.P*).
If Il E Q', Il* E Q', Op E Il but not p e Il*, and if Il* is a
provisional alternative to Il, then we may adjoin p to Il*.
(A.OO+) 'If Il e Q', Il* E Q', Op E Il but not Op E Il* and if Il* is a
provisional alternative to Il, then we may adjoin Op to Il*.
(A.O)rest If Il* E Q', Op E Il*, but not p E IJ.*, then we may adjoin p
to Il* provided that Il* is a provisional alternative to some
member Il of Q'.
(A.o*) If Il e Q' and Op E Il but not P E Il* for any provisional
alternative Il* to Il in Q', then we may adjoin {p} to Q' so
as to become a provisional alternative to Il.
The same remark pertains to (A.o*) as to (A.P*).
How can these rules be employed to show that A. is not satisfiable? Let
us suppose that we have arrived at some provisional model system, and
suppose further that some member Il of Q' violates (C.,....,). Then it is
obvious from the definition of satisfiability that we have no hope of
reaching a model set by adjoining new formulae to the members of Q'.
The violation is there to stay.
SOME MAIN PROBLEMS OF DEONTIC LOGIC 77
14. The literature of deontic logic offers instructive and amusing exam-
ples of such fallacies. For instance, the following plausible-looking prin-
ciple has been put forward:
(14) If we are obliged to do A, then if our doing A implies that
we ought to do B, we are obliged to do B.
This certainly looks like a "quite plain truth" of logic, and it was taken
to be one by A. N. Prior in the first edition of his Formal Logic. He
formulated it essentially as follows:
(15) (Op & (p ~ Oq)) ~ Oq.
Whatever obviousness may seem to accrue to (14) belongs in fact to the
validity of the corresponding deontic consequence, i.e. to the validity of
(16) O«Op & (p ~ Oq)) ~ Oq).
This is shown by the fact that (15) is not valid on the assumptions we
have made, i.e. that its negation
(17) 0p&(-pAOq)&P-q
is satisfiable. This satisfiability is shown by the model system which
consists of the following two model sets:
(18) {Op,(- p v Oq),P - q, - p,(Op&(- p v Oq)&P - q)}
(19) top, p, - q}
of which the latter is assumed to be a deontic alternative to the former.
That this set of sets of sentences satisfies the defining conditions of-a
model system can be verified by inspection. (For simplicity, it is assumed
here that (C.&) is extended so as to apply also to conjunctions with more
than two members.)
What is even more interesting, our model system brings out the reason
why (15) is not valid. Model set (18), which may be viewed as representing
80 JAAKKO HINTIKKA
the actual world, contains both the sentence Op and the sentence '" p.
Using the terms employed in (14), this means that we can escape our
obligation to do B simply by failing to carry out the earlier obligation to
do A. There is nothing logically impossible or even logically awkward
about such a course of events: lamentable as it may be, many of our
obligations in fact remain unfulfilled. This 'escape' can only be prevented
by requiring that all our obligations be fulfilled. But to require this is in
effect to understand (14) as expressing a deontic consequence rather than
a logical one, i.e. as expressing the validity of (16) and not of (15).
That we really have a deontic consequence here can be demonstrated
by showing that the negation of (16), i.e.
(20) p(Op & ('" p v Oq) & P '" q),
cannot occur in any member of model system. This can be accomplished
by trying to build such a model system step by step by means of our rules
of analysis, and by showing that all ways of doing so lead to a violation
of (C. "'). The following argument serves to establish this and incidentally
also serves to illustrate the way in which the rules of analysis can be used
to establish validity:
We shall use one and the same letter D for the ever-growing approxi-
mations to the model system to be built (i.e. for successive provisional
model systems), and the lower case greek letters p, v, ... for their ever-
growing members. At each stage, D, p, v, ... contain only those members
they have been specified to have in the arguments so far. The first line of
our reductive argument shows what formula we are constructing a model
system for:
(21) p(Op & ('" p v Oq) & P '" q) E P. ED.
(22) Op & ('" p v Oq) & P '" q EVE Q.
Here (26) is obtained from (23) by (A.O)rest. This rule is applicable because
v is a provisional deontic alternative to JI..
From (24) we obtain by (A. v)
(27) - pE V
or
(28) Oq E V.
But (26) and (27) violate (C. -). Hence we can restrict our attention to
(28). But (28) and (25) likewise violate (C. -), showing (in virtue of the
soundness of our method) the inconsistency of (20). If you are hesitant
about applying (C. -) to statements containing deontic operators, you
may continue the argument as follows:
(29) - q E e
from (25) by means of (A.P*). Here eis a provisional deontic alternative
to v in Q. But then we have
(30) q E e
from (28) by means of (A.O*). Here (29) and (30) violate once again
(C. -), this time as applied to simpler - possible atomic - statements.
This negative outcome of all the different ways of constructing a world
system for (21) shows the desired validity of (16). In our approach to
deontic logic, the situation is thus the one I claimed it to be: (14) has to
be interpreted as expressing a deontic rather than a logical consequence.
Nor is the possibility of showing this restricted in any way by the peculiar-
ities of our approach. In Prior's old system, the impossibility of adopting
(15) as a valid logical principle is shown by the unnatural consequences
of an attempted adoption. Prior dedu<:cs from this assumption a version
of the so-called first paradox of commitment. This is not the only awk-
ward consequence, however, nor the most awkward one. By substituting
(p&- p) for q in (15) and by noticing the disprovability of O(p&- p)
in most systems in deontic logic, including Prior's, we can readily deduce
from (15) the striking theorem
(31) Op => p.
This says that all obligations are in fact fulfilled, i.e. that we are dealing
82 1AAKKO HINTIKKA
one can find in the literature can scarcely be said to have resulted in any
kind of consensus. If we express the concept of possibility by the modal
operator M (and the associated concept of necessity by the operator N),
it lies close at hand to think of the problem as being concerned with the
logical status of statements of form
(35) Op ::> Mp.
It is easily seen that on the assumptions so far made (together with
certain unproblematic assumptions concerning the notions of necessity
and possibility alone, unrelated to deontic notions) (35) is not valid. Of
course it can be made valid by adopting some further principles concerning
the interplay between the notions of possibility and necessity and deontic
notions. The merits and demerits of such principles would require a
longer discussion than can be undertaken here. It nevertheless seems safe
to say that no obvious and uncontroversial principle is forthcoming on
the level at which we are here moving to restore the validity of (35). In
a context of a logical discussion, it therefore seems advisable not to try
to salvage the 'ought implies can' principle by means of additional
assumptions.
It is perhaps worth emphasizing that a particularly forceful type of
argu1Dent for some versions of the principle is inapplicable here. It is
often said that 'ought implies can' because a man cannot be blamed for
not doing what he cannot do. And if he cannot be blamed for not doing
something, he cannot be under an obligation to do it. Hence his being
under such an obligation presupposes that he can fulfill it.
Whatever the merits of this line of argument are, it is inapplicable here.
What we are dealing with in the present paper are impersonal norms
rather than duties or obligations that pertain to some particular person.
(Whether, and if so how, the latter can be analyzed in terms of the former
is a question which will not be taken up here.) But if so, the argument just
sketched for the 'ought implies can' principle falls outside the scope of
the present paper, too, for it trades essentially in obligations of some
particular person. (Only by so doing can the crucial notion of blame be
brought in. For no one in particular can be blamed for not fulfilling an
'impersonal' norm of the kind we are here dealing with.)
Hence it may appear that very little can be said about the 'sollen-
konnen' principle here. One simple point can nevertheless be made.
SOME MAIN PROBLEMS OF DEONTIC LOGIC 8S
Whatever the status of (3S) is, there is no problem about the status of a
closely related sentence. Even if it is the case that Mp is not a logical
consequence of Op, it is without any doubt a deontic consequence of the
latter. In other words, sentences of the form
(36) O(Op ~ Mp)
are valid already in virtue of the assumptions we have made, plus one
unproblematic assumption concerning the logical behaviour of the con-
cepts of necessity and possibility.
This can be shown by means of our rules of analysis. An argument to
this effect can be carried out by trying to imbed the negation of (36) in
a member of a model system. As in our earlier argument, we shall use
a for the successive provisional model systems obtained in the course of
the arguments, and I" for the successive approximations of that member
of 0 in which we want the negation of (36) to occur. The argument may
proceed as follows:
(37) P(Op&N-p)el"ea.
This just says what we want to construct our model system for. (In
bringing the negation of (36) to the form here displayed, we have made
use of the equivalence of "" Mp with N ,.., p.)
(38) (Op&N-p)ev.
This is obtained from (37) by means of (A.P*). Here v is a provisional
deontic alternative to I" in D.
(39) Op e v (from (38) by (A.&)l)
(40) N - p e v (from (38) by (A.&)2)
(41) pev.
This is obtained from (39) by means of (A.O)rnv which is applicable
because v is a provisional deontic alternative to 1".
(42) - p e v.
This is obtained from (40) by the scarcely disputable principle that what-
ever is necessarily true (in a given possible world) is true (there). But (41)
and (42) contradict (C. "'), completing the desired argument, and thus
86 JAAKKO HINTIKKA
Much ink has been spilled in discussing th~ relative merits of these two
explications. It has been spilled in vain, for the conclusion seems to me
inescapable that our commonplace notion of commitment is intrinsically
ambiguous between the two renderings (43) and (44) (plus, possibly, still
others).
Our semantical insights enable us to appreciate the difference between
(43) and (44). The former reconstruction in effect assimilates, in the
special use in which (43) is logically true, the notion of commitment to our
earlier notion of deontic consequence. On this interpretation p commits
us to q if it is impossible to realize p in a 'deontically perfect world' with-
out realizing q, too. Since it has already been seen that our logical
intuitions in the area of normative concepts often in effect pertain to
88 JAAKKO HINTIKKA
17. This does not show, however, that (43) is always what people's
informal verbal statements about commitments presuppose even in
perfectly normal circumstances. In fact, there are good general reasons
for thinking that often (43) is not the intended interpretation. For one
thing, from (43) together with a purely factual statement no unconditional
statements of obligation follows. For instance, p and O(p::> q) do not
imply Oq. In this sense, commitments of the kind (43) do not admit
'detachment'. Yet on some occasions we certainly consider ourselves
justified to carry out such a detachment and to announce, on the basis
of a fact and a commitment, a definite non-conditional obligation.
A commitment for which this is possible must have something like the
force of (44) rather than (43). Whenever an actual obligation follows from
a commitment plus certain facts, some rec')nstruction along the lines of
(44) rather than of (43) is thus presupposed. Such cases seem in fact to
be quite common.
Objections have been made to (44) as an interpretation of the notion
of commitment. For instance, it has been alleged that on this interpre-
tation the realization of whatever is not in fact realized 'commits' one
to everything, for
(47) ,.., p ::> (p ::> Oq)
is valid.
The fact is that what creates the appearance of a paradox here is not so
much the idea on which (44) is based as rather the desire to have some
90 JAAKKO HINTIKKA
18. The differences between (43) and (44) - as well as the reasons for using
both of them as alternative explications of the notion of commitment -
are illustrated by conflicts of duty. Such a conflict may e.g. result from a
promise. I give an honest promise (let this act be described by p) to bring
it about that q (say, have a cup of coffee with you). Unknownst to me,
however, my father has fallen ill, which creates an obligation to see him
that overrules my earlier promise. It seems to me that moral philosophers
have felt somewhat uncomfortable in discussing this kind of situation,
and perhaps one can also see why. For the fact that the obligation created
by my promise is overruled means that it is false to say simpliciter that
Oq, i.e. that I. am obliged to fulfill the promise. (By the same token, the
commitment involved in my promise cannot be construed as (44)). Yet
it is clear that not everything is morally all right if I have to break my
promise, however firmly this particular course of action may be pre-
scribed to me by the norms I abide by. I have somehow done something
wrong. This 'moral failure' is what easily makes one hesitant to say that
in such a case there is no absolute duty to keep the promise.
Our distinction between (43) and (44) enables us to see precisely what
goes wrong in such a case. It is obviously and clearly true, even in the
case of a promise overruled, that in a deontical/y perfect world such a
promise cannot be given without keeping it. In such a world, p cannot be
realized without bringing it about that q. Even if the act of promising
does not give rise to an actual duty to keep the promise (e.g. because of
other duties), it none the less remains true that in this sense giving a
promise commits one to keeping it.
The sense of commitment involved here is clearly (43). Thus it may be
said that we need sense (43) to account for the possibility that a perfectly
SOME MAIN PROBLEMS OF DEONTIC LOGIC 91
19. By this time the reader has - hopefully - begun to appreciate the
difference between (43) and (44). At the same time, an especially attentive
reader may also have had an experience of deja vu - of recognizing some-
thing he recalls from the literature of moral philosophy.
In fact, I have already slipped a few times into a bit of conventional
jargon by speaking in connection with (44) of actual or absolute obli-
gations. To make uninhibited use of this jargon, the contrast between (43)
and (44) is essentially that between prima facie duties (obligations) and
actual (absolute or overall) duties or obligations. 2
It is in fact obvious that the situations we considered for the purpose
of illustrating the difference between (43) and (44) are of the same kind
as those which the perpetrators of the traditional dichotomy have used
as paradigm cases of the contrast between prima facie duty and actual
duty. The main problem to which they have addressed themselves is
likewise admirably accounted for by our distinction. This problem is the
question as to how an obligation can be overruled and yet remain - in
some perfectly good sense - a genuine obligation. Our answer to this
question was already given. It is now seen to admit of a formulation in
terms of the traditional distinction.
In order to obtain an explicit reconstruction of the distinction prima
facie obligation vs. actual obligation, let us consider some set of normative
principles whose conjunction is n. (The sentences formulating these
principles may be of the form Oq, but they may also exemplify such more
complex forms as (43) or (44).) Let us also assume that we have as our
factual premises a set of descriptive statements whose conjunction is p.
Then we shall say that on the basis of the set of norms n, q is a prima
facie obligation if and only if
(49) (n & p) ~ Oq
92 JAAKKO HINTIKKA
is valid, i.e. if and only if Oq is a logical consequence of ('I &p). Thus the
distinction between prima facie obligations and actual obligations is
closely related to the distinction between the notions of deontic conse-
quence and logical consequence. The ambiguity of our intuitions vis-a-vis
this distinction is probably the major reason why the notions of prima
facie obligation and actual obligation have been distinguished so late
and confused so often. (For a discussion of an example of such confusion,
see Sections 22 and 23 of the present essay.)
A logician is amused to find that an important philosophical distinction
once again turns out to be based on an operator-switch, i.e. on the
order of two different logical operations, in the case at hand, between
o and ::J.
Sir David Ross, who more than anyone else has been instrumental in
introducing the concept of prima facie duty (obligation) into contempo-
rary moral philosophy, uses the term absolute duty instead of actual duty
for its contrary. This is not incompatible with our reconstruction. The
fact that in (48) the deontic operator 0 governs a conditional (if-then)
sentence shows in what sense prima facie obligations are in our view non-
absolute or conditional.
If n does not contain any normative notions, there is a prima facie
obligation tha.t q if and only if q follows logically from the non-normative
premise (n&p). This fact throws some light on the notion of a technical
norm and on its relation to other kinds of norms.
wrong in such cases. We have already seen, however, that this feeling is
sufficiently accounted for by pointing out the precise sense in which a
primafacie duty is violated: something takes place that would not happen
in a deontically perfect world. Indeed the actual breach of morality which
takes place in such cases is typically different from a failure to satisfy a
prima facie obligation. For instance, in the case of (say) promising the
only conclusion we can detach (the only actual duty we can infer) is the
actual duty not to give the kind of promise that will be overruled by
other obligations, assuming that the duty to fulfill a promise is a prima
facie one. In other words: although fromp and O(p::.> q) we cannot infer
Oq, we can from p, 0 (p ::.> q), and "" Oq infer 0 "" p.
Toward the end of this essay we shall meet one more instance of a
philosopher's failure to see how easily prima facie obligations can obtain
without any corresponding actual (absolute) obligation obtaining, and
without anything going wrong with our logic.
A reason for the importance of prima facie obligations follows from
our earlier remark that we are likely to have firmer views concerning
how things ought to be, i.e. what a deontica1ly perfect world looks like,
than concerning the multiple interrelations of actual duties. For what
prima facie obligations specify is precisely what happens in deontically
perfect worlds.
A bonus we obtain as a by-product of our reconstruction of the distinc-
tion between prima facie duties and actual duties is a handy terminology
for the distinction between the two kinds of commitment (43) and (44)
which was discussed above at length. The former may be called - and
from now on will be called - primafacie commitments and the latter actual
or absolute commitments.
act of promising together with the analytical (Searle uses the term
'tautological') premise that promises ought to be kept it follows that
there is an obligation to keep the promise in question (an 'ought'). In
short, an ought follows from an is plus a tautological and hence empty
additional premise.
Let p be a statement to the effect that a certain particular promise is
given and let q state that this particular promise is kept. We can relate
Searle's discussion to our own earlier discussion by expressing his
'tautological' second premise by saying that giving the promise in question
commits one to keeping it. Thus Searle's 'derivation of ought from is' is
(on our oversimplified reconstruction) of the form
P
(50) p commits one to q
Oq
P
(50*) O(p::::> q)
Oq
and
p
(50**) p::::> Oq.
Oq
This distinction between two senses of (50) corresponds neatly to the two
senses of Searle's locution 'placing oneself under an obligation'. In (50*)
SOME MAIN PROBLEMS OF DEONTIC LOGIC 95
p does not contain any normative elements, then the ceteris paribw
condition cp will be at least partly normative.
We might thus represent schematically the three interpretations of
(50**) which we have considered as follows:
p (factual)
(51) p => Oq (non-analytical)
Oq
p (normative)
(51 *) p => Oq (analytical)
Oq
(p & cp) (normative)
(51**) (p & cp) => Oq (analytical)
Oq
Although all these three represent logically valid inferences, they fail
to provide us with a 'derivation of ought from is' in the intended sense.
The specious plausibility of assuming, in the third line of interpretation
just mentioned, that the ceteris parihw condition cp can be taken to be
factual is witnessed by Searle's adherence (essentially) to this line of
defense. He formulates the second premise as follows: "All those who
[promise, i.e.] place themselves under an obligation are, other things
being equal, under an obligation." He explains the need of the qualifying
clause here by saying that "we need the ceteris paribus clause to eliminate
the possibility that something extraneous to the relation of 'obligation'
to 'ought' might interfere." The interfering factors that Searle here labels
'extraneous' include competing stronger obligations, which the ceteris
paribus clause must also eliminate in order to serve its purpose. But they
cannot be ruled out by means of purely factual assumptions.
Searle nevertheless strives to maintain that "there is nothing necessarily
evaluative about the ceteris paribus conditions." His argument hinges
on the observations that "an evaluation [of the competing obligations]
is not necessary in every case" and that "unless we have some reason to
the contrary, the ceteris paribw condition is satisfied, and the question
whether he ought to do it is settled by saying 'he promised'." This argu-
ment has no force, however. It is true that in some cases no intervention
98 lAAKKO HINTIKKA
takes place, i.e. that in some normative situations there are no conflicting
obligations. But to state that this is the case is to make a normative
statement. Saying that t is deontically neutral (,..., Ot &,..., 0 ,..., t) is as
much a normative statement as saying that it is obligatory or forbidden.
Likewise, to say that a prima facie obligation is not overruled by others
(and that the question of actual duty can be decided in the way Searle
says) is to make a normative statement, however negative. And to try to
tie the need of evaluation to the question whether counter-arguments
have in fact been presented, or whether we (actually?) have 'reasons to
the contrary', as Searle appears to do, is of course beside the point. The
question is what is implied by the norms one has accepted, not what
arguments have actually been put forward or what reasons one actually
has.
23. This by no means exhausts the interest of Searle's clever paper, nor
even the different types of argument he is considering. A closer exami-
nation of these alternative arguments would uncover flaws in them
similar to those we have already discussed. 4 However, our sole purpose
here is to illustrate such important distinctions as (43)-(44) by means of
Searle's main argument, which does not motivate a discussion of the
further details of his paper.
One can nevertheless use our distinctions also to emphasize the extent
to which Searle is perfectly right. If 0 (p ::> q) is a principle of one's
normative system, however analytical, then one can after all infer from
p that there is a prima facie obligation to bring it about that q. (To see
this, put O(p::> q) for n in (48) and try to assume that its negation is
satisfiable.) Hence Searle is right in a rather striking sense. He has in
effect pointed out that from an 'is' and from an analytical principle one
may legitimately derive a perfectly genuine obligation, viz. a prima facie
obligation. An 'ought' does follow from an 'is', albeit only a prima facie
'ought'. This observation becomes all the more important in the light of
our earlier observation that such a prima facie 'ought' is often what our
intuitions are all about anyway.
From this point of view, the basic flaw of Searle's paper does not
consist so much in putting forward a fallacious argument as in failing to
spell out the sense in which his (correct) conclusion is to be understood.
He is calling our attention to a perfectly legitimate relation of deontic
SOME MAIN PROBLEMS OF DEONTIC LOGIC 99
25. It was indicated earlier that our conditions and rules concerning
quantifiers are in need of modifications. In the context of the usual
modal operators, this need is pointed out in my paper, 'Modality and
Quantification', Theoria 27 (1961) 119-128 (reprinted with additions in
Modelslor Modalities, D. Reidel, Dordrecht, 1969). For details, it suffices
to refer the reader to these other discussions. The main point is clear
enough in any case. The conditions (C. U) and (C.E) obviously presuppose
that each individual whose name occurs in the members of p exists in the
possible world described by p. Now an individual existing in one world
need not exist in others. Hence rules which for the first time import a
new free individual variable into (an approximation toward) a model set
are suspect, and have to be modified so as to rule out illicit impor-
tation.
By inspection, it is seen that there are two such rules, viz. (A.O+) and
(A.OO+). They will have to be changed as follows:
(A.O*) Like (A.O+) except that it is also required that all the free
individual variables of p occur in the members of p*.
(A.OO*) Like (A.OO+) except that it is also required that all the free
individual variables of p occur in the members of p*.
Academy of Finland.
NOTES
1 The main difference between the distinction (43)-(45) and the earlier distinction,
deontic consequence vs. logical consequence, is of course that neither (43) nor (44)
has to be true for logical (conceptual reasons, whereas the latter distinction dealt with
two kinds of logical (conceptual) connections between statements.
2 The primary sources of this distinction in recent moral philosophy are the writings
of Sir David Ross, especially The Right and the Good, Clarendon Press, Oxford, 1930,
and The Foundations of Ethics, Clarendon Press, Oxford, 1939.
3 John R. Searle, 'How to Derive Ought from Is', Philosophical Review 73 (1964)
43-58, reprinted in Jerry H. Gill (ed.), Philosophy To-Day no. 1, The Macmillan Co.,
New York, 1968, pp. 218-235, and in Philippa Foot (ed.), Theories of Ethics, Oxford
Readings in Philosophy, Oxford University Press, Oxford, 1968, pp. 101-114.
4 A case in point is the following: Searle says that the kind of criticism I just presented
is in any case inconclusive, "for we can always rewrite the relevant steps ... so that
they include the ceteris paribus clause as part of the conclusion."
104 JAAKKO HINTIKKA
This sounds fine. However, everything depends on the precise way in which the
incorporation of the ceteris paribus condition in the conclusion IS supposed to be
accomplished. There are two possibilities which yield essentially the following putative
arguments:
p
(*) p O(cp ::> p)
::>
O(cp ::> p)
P
(**) p ::>(cp ::> Op)
(cp::> Oq)
Now the second premise of (*) is obviously false. Surely it does not follow from the
fact that a promise is given in the actual world that in all deontically perfect worlds cp
is followed by q. Hence (*) must be ruled out.
The only way to avoid this conclusion is to make cp ~ p a logical (analytical) truth.
This, however, deprives (*) of all relevance for Searle's purpose.
In (**), the conclusion is a conditional with tacitly normative antecedent and an
explicitly normative consequent. That such a statement follows logically from a factual
statement together with an analytically true premise has no implications whatsoever
for the is-ought distinction, any more than (say) the logical truth of Oq ::> Oq has.
5 In his new book, Speech Acts, Cambridge University Press, Cambridge, 1969, p. 181,
Searle now distinguishes between two kinds of obligations, exemplified by
All things considered, Jones ought to pay Smith five dollars
and
As regards his obligation to pay Smith five dollars, Jones ought to pay
Smith five dollars,
respectively. He says that only obligations of the latter type, not of the former, can be
derived from an 'is'. This distinction comes very close to our distinction between
absolute obligations and prima facie ones. Searle fails to spell out, however, precisely
what is involved in the latter. I claim that when this is done, the limitations of Searle's
argument become patent. A derivation of a prima facie ought from an 'is' does not
violate the fact-norm dichotomy, correctly understood.
GEORG HENRIK VON WRIGHT
R. Hilpinen (ed.) , Deontic Logic: Introductory and Sy.tematic Reading., 105-120. All right. re.erved.
Copyright © 1970 by D. Reidel Publishing Company, Dordrecht-Holland.
106 GEORG HENRIK VON WRIGHT
atomic O-expressions and their molecular complexes. (T1:le rules for the
use of brackets in formulae I shall not mention here. Which they are
should be obvious from the presentation.)
The formula 'OA' may be read as follows: one ought to see to it that A.
And similarly for other formulae. For example: If 'A' describes the state
of affairs that the window is open, 'OA' says that one ought to see to it
that the window is open. - The reading of 'OA' as one ought to do A
I do no longer regard as fully correct.
The system has two axioms:
(AI) ,..., (OA & 0,..., A).
III
For example: Assume that it is our moral duty not to hurt anybody's
feelings. Assume that we have hurt somebody's feelings. Then our duty
could be to apologize. This, I understand, would be a contrary-to-duty
imperative in Chisholm's sense.
It is obvious that contrary-to-duty imperatives play an important rOle
in the moral life of man, and in law. 'Rules of reparation', i.e. norms
concerning how to make good some damage caused or some evil done,
are often of this kind. It is a definite merit of Professor Chisholm's to
have drawn attention to this type of imperative in the context of deontic
logic.
IV
VI
left and right of 'f' are schemas for descriptions of two ..,ossible generic
states of affairs. The description to the left tells us, how the world ought
to be, when it is as the description to the right says that it is.
Sometimes the world is as it ought to be. It is thoroughly meaningful
to make it a duty that o (A/A). The duty to see to it that A when it is the
case that A requires us to take heed that the state of affairs in question
does not disappear.
Not always, however, is the world as it ought to be. Then the duty may
be that 0 (A/ '" A). The duty to see to it that A when this is not the case
requires us to take care that the state of affairs in question comes to be.
It should be clear from these considerations that there is no objection,
from the point of view of interpreting the formulae, to the possibility
that the same variable appears both to the left and to the right of the 'f'
in the same atomic O-expression.
The New System has three axioms:
Bl '" [0 (A/B) & O( '" A/B)].
B2 O(A & B/C) - o (A/C) & o (B/C).
B3 O(A/B v C) - o (A/B) & O(A/C).
The follow~ng example should convince us of the intuitive plausibility
of the third axiom: Suppose we are given the order to see to it that the
window is closed, should it start raining or start to thunder. Obviously
this is equivalent to being given the order to see to it that the window is
closed, should it start raining, and see to it that the window is closed,
should it start to thunder. It is an observation of some interest about the
functioning of ordinary language that the 'or' in the undistributed con-
ditioning clause of the order should become an 'and' when the clause is
being distributed.
The rules of inference of the New System are the same as the rules of
inference of the Old System (RI-R4).
VII
Consider a formula which is (an axiom or) a theorem of the Old System.
Let 'T' stand for an arbitrary tautology formed of the variables A, B, ...
and truth-connectives. Replace each atomic O-expression '0 ( -)' which
A NEW SYSTEM OF DEONTIC LOGIC 111
VIII
IX
Special rules must be given for the cases, when a compound to the left of
'j' in an atomic O-expression is a tautology and when a compound to the
right of 'j' is a contradiction. 8 In the first case the compound has no
perfect conjunctive and in the second case no perfect disjunctive normal
form. The normal forms 'vanish', contain no disjunction and conjunction
respectively.
We first note the following two theorems:
Tl o (AlB) -+ D(A v "" A/B).
'If under some circumstances a certain thing is obligatory, then under
A NEW SYSTEM OF DEONTIe LOGIC 113
then all constituents, if there are any, of the form '0 (T/ U)' and of the
form '0 (S/", T)' must be assigned the value 'true'.
(iii) If some constituent of the form' 0 (T/ S)' or '0 (S/ '" T)' is assigned
the value 'true', then all constituents, if there are any, of the form
'0 (TI '" T)' must be assigned the value 'true'.
Consider the two imperatives expressed by 'O(AIT)' and '0 (BI'" A)'.
The first is a categorical or unconditional order under any circumstances
to see to it that A. The second is a hypothetical or conditional order to
see to it that B, should it be the case that '" A.
It is easily shown that the categorical imperative does not entail the
hypothetical one, but that the two are logically independent. The cate-
gorical order can be expanded as follows: O[(A v B)&(A v'" B)/A &B
v A &'" B v '" A &B v '" A &,.., BJ. The hypothetical order again is
equivalent to 0 [(A v B) & ( '" A v B)/'" A &B v '" A & '" BJ. The distri-
bution-axioms B2 and B3 may be used for splitting up the two expressions
into O-constituents. Then the categorical order becomes a conjunction
of 8 and the hypothetical order a conjunction of 4 O-constituents. Two
of the constituents of the hypothetical order are not among the consti-
tuents of the categorical order. This suffices to show that the two orders
are logically independent.
Shall we say that '0 (B/,.., A)' expresses the contrary-to-duty impera-
tive to see to it that B in case a duty-bound subject has neglected his
primary duty to see to it that A? We shall not say this. For, there is
nothing in the form of the imperative to show that the state of affairs
that '" A has come about as a result of neglect of duty. To show this we
should have to make use of a symbolism which contains names of agents
and has means of distinguishing between various ways in which states
of affairs may come to be. Such a symbolism can be developed, - but we
shall not try to develop it here.
Contrary-to-duty imperatives are a special class of hypothetical im-
peratives. The logical structure of hypothetical imperatives, I shall
maintain, is that of a dyadic obligation-operator which obeys the axioms
BI-B3 of the calculus here called the New System of Deontic Logic.
Neither the formula '0 ( '" A -. B)' of the Old System, nor the 'mixed'
A NEW SYSTEM OF DEONTIC LOGIC 115
XI
The immediate purpose which I had in mind when constructing the New
System was to deal with certain difficulties which had been pointed out
by Professor Chisholm. l l The difficulties in question centered round a
notion for which Chisholm had coined the name "contrary-to-duty im-
peratives". A contrary-to-duty imperative, generally speaking, tells us
what ought to be done when a certain (primary) duty has been neglected.
As shown by Chisholm, certain systems of deontic logic, among them
my Old System, cannot adequately formalize such imperatives. In my
paper I wanted to show that the New System captures the general form
of contrary-to-duty imperatives.
What was said on contrary-to-duty imperatives in my paper remains,
for all I can see, untouched by the objection to the New System which I
shall presently mention and by the amendment of the system which I am
going to propose.
XII
In the New System one can prove certain things which are counter-
intuitive and therefore not wanted in a sound deontic logic. I am much
obliged to Mr. P. Geach for his having brought this to my attention.
Let there be a norm to the effect that one ought to see to it that A is
the case, should it be the case that B. In symbols: O(A/B). By virtue of
the Rule of Extensionality, o (A/B) is equivalent with O(A/B&C v B&
'" C). This, by virtue of B3, is equivalent with O(A/B&C)&O(A/B&
'" C) which entails o (A/B& C). Thus we have proved the formula
o (A/B)-+ O(A/B&C).
116 GEORG HENRIK VON WRIGHT
By similar arguments we prove 0(", AIC)-+ O( '" AlB & C). This is
equivalent with '" 0 ( '" AlB & C) -+ '" 0 ( '" A/C).
By virtue of Bl, we have", [O(A/B&C)&O( '" A/B&C)] which is
equivalent with O(AIB&C)-+ '" O( '" A/B&C).
O(AIB)-+ O(A/B&C) and O(A/B&C)-+ '" O( '" A/B&C) yield to-
gether O(A/B)-+ ",O( '" A/B&C).
O(A/B)-+ '" O( '" A/B&C) and", O( '" A/B&C)-+ '" O( '" A/C) yield
together 0 (A/B) -+ '" 0 C·v A/C).
Herewith it has been proved that, if there is a duty to see to it that A
under circumstances B, then there is no duty to see to it that not-A under
circumstances C. For example: It has been proved that, if there is a duty
to see to it that a certain window is closed should it start raining, then
there cannot be a duty to see to it that the window is open should the
sun be shining. This is manifestly absurd. Generally speaking: From a
duty to see to a certain thing under certain circumstances nothing can
follow logically concerning a duty or not-duty under entirely different,
logically unrelated, circumstances. Least of all should one be able to
prove that there is under those unrelated circumstances a duty of contra-
dictory content.
XIII
Can the root of the trouble be traced and the calculus put right? I think
this is possible.
The source of the trouble, as I see it, resides in axiom Bl. This axiom
was meant to be an analogue of the axiom
Al '" (OA & 0 '" A)
of the Old System of Deontic Logic, i.e. of the monadic calculus of un-
conditional or categorical or absolute norms (duties).
The analogue, however, was falsely conceived.
An expression formed of the letter '0' in front of a disjunction of single
variables and/or their negations will be called a (normal) monadic 0-
constituent. n variables determine 2" monadic O-constituents. We shall
also call the constituents the O-units of the deontic space determined by
n variables (for generic states of affairs).
One variable A determines a deontic space of two O-units, OA and
0", A. Axiom Al thus says that the two deontic O-units of the deontic
A NEW SYSTEM OF DEONTIC LOGIC 117
The New System of Deontic Logic, thus amended, caters for a logical
possibility, to which the system in its original form cannot do justice.
This is the possibility of conflicting duties (obligations).
We shall say that duties conflict, if they require the doing of conjunc-
tively impossible actions under the same circumstances. For example:
a duty to see to it that A when B and a duty to see to it that '" A when B
are a pair of conflicting duties. In symbols: 0 (AI B) and 0 ( '" AlB) are
conflicting. According to the original version of the system, they cannot
both be duties. This was excluded by axiom Bl. According to this system,
a (genuine) conflict of duties was therefore a logical impossibility. This
it obviously is not. One thing which the derivation of the absurdity in
Section XII shows, is the necessity of allowing for the possibility of
conflicting duties in a sane system of conditional norms.
The axiom Al of the Old System excludes conflicting duties too. Does
it mean that the system is not sound? It does not mean this. For, it may
truly be regarded as a logical impossibility that absolute (categorical, un-
conditional) duties or obligations should conflict. (Unconditional duties
under different laws or systems of norm may, of course, conflict, in the
sense that they impose logically contradictory or contrary demands on
an agent. Such cases, however, are logically uninteresting and should
better not be regarded as genuine 'conflicts of duties' at all.)
It is logically absurd that a man should have conflicting duties whatever
his circumstances are. Conflicting duties arise only under special circum-
&tances. Such circumstances constitute what may be called a (moral)
predicament.
Of particular interest both from a logical and from an ethical point of
view is the predicament which arises when a man promises to do the
A NEW SYSTEM OF DEONTIe LOGIC 119
Academy of Finland
NOTES
• Sections I-X of this paper are a reprint of 'A New System of Deontic Logic',
Danish Yearbook 0/ Philosophy 1 (1964) 173-182; Sections XI-XIV are a reprint of
sections 2-5 of 'A Correction to a New System of Deontic Logic', Danish Yearbook
o/Philosophy 2 (1965) 103-107. Reprinted by permission of the author and the publisher.
1 Roderick M. Chisholm, 'Contrary-to-duty Imperatives and Deontic Logic', Analysis
24 (1963) 33-36.
2 G. H. von Wright, 'Deontic Logic', Mind 60 (1951) 1-15.
S There is a (fragmentary) discussion of these notions in my book Norm and Action,
Routledge and Kegan Paul, London, 1963, pp. 22-27.
4 Chisholm, ibid., p. 34: "It is clear that we must use instead a conditional with an
obligatory consequent and tell him 'If you steal then it is obligatory that you return
the money'."
5 I say "seems to entail" because I am not quite certain whether 'A' and '"" A' can
be correctly interpreted as saying that the agent has fulfilled, respectively neglected,
his duty to see to it that A. If they cannot be thus interpreted, however, then another
objection to Chisholm's suggestion seems to be forthcoming. This new objection is
that' "" A -+ 0 B' does not show that the state of affairs described by '"" A' is supposed
to have come about as a result of neglect of duty - and therefore cannot express a
contrary-to-duty imperative. (Cf. below Section X.)
6 Chisholm, ibid., p. 34: "We may also wish to tell him, if the need arises, 'It is not
true that, if you steal then it is obligatory that you steal again'."
7 Ibid., pp. 34-35.
8 I am indebted to Dr. Lars Svenonius for his having drawn my attention to a defect
in an earlier attempt to formulate these rules.
9 Because of R3 we may regard the contradiction to the right of 'j' as the negation of
the very tautology to the left of 'j' .
10 The view that ',,4 -+ OB' is the logical form of a hypothetical norm has been inter-
estingly argued by Professor Erik Stenius in his paper 'The Principles of a Logic of
Normative Systems' in Proceedings 0/ a Colloquium on Modal and Many- Valued Logics,
Acta Philosophica Fennica 16 (1963), esp. pp. 256-260. It seems to me, however, that
120 GEORG HENRIK VON WRIGHT
Stenius's view is not capable of doing full justice to the logical peculiarities of this type
of norm.
11 Roderick M. Chisholm, 'Contrary-to-duty Imperatives and Deontic Logic', Analysis
24 (1963) 33-36.
12 For the proof of an analogous principle in the monadic calculus see my paper
'Deontic Logic', Mind 60 (1951), p. 13.
BENGT HANSSON
I. INTRODUCTION
I assume the existence of a basis logic (henceforth BL) which may be the
propositional calculus or the first-order predicate calculus with or without
constants, or some related system. It is essential that valuations and
validity can be defined in the usual way and that BL is complete in the
sense that every valid formula is a theorem. I also assume that it is well-
known what constitutes a well-formed formula and a theorem of BL
R. Hllpinen (ed.). Deontic Logic: Introductory and Systematic Readings. 121-147. All rights reserved.
Copyright © 1970 by D. Reidel Publishing Company. Dordrecht-Holland.
122 BENGT HANSSON
mulas. But a phrase like 'it is obligatory that p' is generally not considered
to be a true or false statement. On the other hand a deontic statement is
not simply an imperative (although some philosophers hold the view that
there is a sense in which it is equivalent to one), because one can point
out to a person that he ought to do so-and-so without actually telling
him to do it. I will here take the view that deontic statements (formulas
of SDL) are descriptive, that they describe what is obligatory, forbidden
and permitted respectively, according to some (undetermined) system of
norms or moral or legal theory. The proper reading of e.g. 'P' should
then be: 'Norm-system X permits that .. .'. The deontic axioms which
will be discussed later, then, do not have the status of logical truths, but
they express properties of the norm-system used. Those who are attracted
by the axioms may then, if they so want, regard them as criteria of ratio-
nality or of inner coherence of norm-systems or moral or legal theories.
This descriptive interpretation has one advantage in addition to
making formulas of SDL into propositions: it makes clear that deontic
logic is a tool of meta-ethics and not a part of ethics proper.
The interpretation could be supplemented with a reference to the
person or class of persons to which the norms are directed. If this is done,
I will presuppose that the reference is held constant throughout one and
the same context.
In giving this interpretation I do not claim that it is the only one which
is good or interesting, only that it provides a ground for discussing the
philosophical implications of certain results about the structure of von
Wright-type deontic logic. Discussions based on other interpretations
may yield interesting results and I do not '.'{ish to exclude them.
confusion. But the reader should be aware that the use oftJ-.e term involves
no ontological commitments.
If/is any formula of BL, let T(f) be the set of possible worlds in which
I is given the value truth. Let t be the set of all possible worlds. "'" I is
given the value true in exactly those possible worlds where I is not true,
i.e., T(-/)=t-T(f). It is also obvious that T(p&q)=T(p)rtT(q)
and T(p v q)=T(p)uT(q).
IfI is a theorem of BL we know that T(f) = t. And if T(/) = t we know
that I is a theorem because BL is complete. The completeness theorem
for BL also says that I and g are provably equivalent if and only if
T(f)=T(g). So if one is just concerned about the content of a formula
and not about its form one could just as well talk about the set T(f)
instead off I will do so because one can always replace a formula by a
provable equivalent in von Wright-type deontic logics. I will further drop
the function symbol T, so that J, g etc. will be treated directly as sets.
Since negation corresponds to complement, conjunction to intersection
and disjunction to union, it does not matter which set of concepts one
uses. I will use t as short for theorem (a tautology) and k as short for
any formula which is the negation of a theorem (a contradiction). k then
is simply the empty set.
If I and g are formulas, then 1--+ g is provable in BL if and only if
I r;;;. g. Note that the lormula 1--+ g is a set (viz. "'" lug), but the fact
that I --+ g is provable is the fact that I is a subset of g.
VI. DESCRIPTIONS
In asserting a formula one really says that the real world is one of the
possible worlds which are elements of the formula, or, equivalently, that
some possibl~ worlds (viz. those in the complement of the formula) are
ruled out as not being actual. We may say that a formula is a description
of the world, but only a partial one, because we can never identify a
possible world as the real one only on the evidence of a single formula.
The information carried by a formula is simply that the range of possible
worlds among which we have to look for the real one is narrowed down
a bit. A complete description would be one which tells us exactly which
set the real world is, i.e. a set containing only one element. But no one-
element set is a formula. What other kind of descriptions do we have
than those provided by formulas? Sometimes we not only assert a single
formula, but a whole class of formulas at the same time. Such a descrip-
tion is very like a theory, because if asked we would probably also assert
finite conjunctions and logical consequences of the formulas in the class.
Therefore such a description can be identified with an assertion of the
basis of the theory generated. This is a slight extension of the use of
'assert'; to assert a set will mean to assert that the real world is one of
the possible worlds in that set.
AN ANALYSIS OF SOME DEONTIC LOGICS 127
Among the descriptions of this new kind we find the complete de-
scriptions, the singleton sets, because every such set is a theory-set (as
can easily be proved).
But we can think of still more general descriptions. We can, e.g., assert
that at least one of the formulas in a certain class is true. This assertion
can be represented by the union of the formulas and this set need neither
be a formula nor a theory-set. And we may have an infinite class of
theories and assert that at least one of them is true (i.e., that in a certain
set of classes of formulas, there is at least one in which every formula is
true), this assertion being represented by the union of the bases, which
again may be a new kind of set. And we may go on and claim that every
Borel set over the set of formulas is a description.
All these kinds of descriptions have only one thing in common: they
are (in a more or less complicated way) generated from the set of for-
mulas, and therefore expressible (admitted in a very general sense) in the
metalanguage of BL.
But the most general kind of description is an arbitrary subset of t.
It provides the information that possible worlds outside that set are in
fact not actual. But for these descriptions we do not in general have any
connection with the metalanguage of BL, so they are in a certain sense
inexpressible. Since the cardinality of the class of formulas is strictly less
than that of the class of theory-sets, which in turn is less than that of all
subsets of t, both formulas and theory-sets are indeed scarce among the
subsets of t. So there is a lot to be said which cannot be said in formulas
or theories. But in one respect theory-sets are nevertheless sufficient, viz.
as regards the class of formulas one is prepared to assert given a certain
description of the world. Such a class should have the structure of a
theory and is therefore fully determined by its basis, a theory-set.
Now when the structure of BL has been discussed at some length, the
structure of SDL will be revealed quite easily. The following two axioms
are suggested by von Wright in [15] and [19]:
Al O(pnq)++Op&Oq
A2 '" O(k)
128 BENGT HANSSON
Now, what does 'P' mean in terms of ideal worlds? Plmeans '" 0 "'" f,
i.e. "'" lis not obligatory, i.e. the basis is not a subset of "'" f, i.e. the basis
and I have a non-empty intersection. Something is permitted, then, if
and only if it does not exclude all ideal worlds. 'Forbidden', accordingly,
is to be interpreted 'does exclude every ideal world'.
Permissions in the sense of SOL must be distinguished from rights in
the sense of claims. Permission to do I means in SOL 'if you do f, you
will not be blamed; you have not broken any rules'. A right normally
entails not only permission, but also obligations for other persons to
refrain from bringing about,..., f. But in SOL it is quite possible that some
person is permitted to do I and another to do '" J, and he who had the
permission to do I must not complain if somebody else did ,..., I thus
making it impossible for him to use his permission. P is then a weak kind
of permission in SOL, but the definition of P in terms of 0 is not essential
to the system. The axioms are formulated in 0 and one may regard SOL
as a system exclusively about obligations if one is not satisfied with the
weak senses of permission.
The case when the basis is a one-element set deserves some special
attention. Then 'having the basis as a subset' is tantamount to 'having
non-empty intersection with the basis' and therefore everything permitted
is also obligatory and vice versa. The norm system or moral or legal
theory which has such a basis leaves no room for free choice without
breaking the rules; for every f, eitherf or ,.., fis obligatory. This situation
arises when the norm system is too detailed: since every change in the
ideal world makes it non-ideal, every formula expresses something
morally relevant. But if we assign meaning to the formulas in such a way
as to cover all aspects of the world, then some formula will mean 'Jones
buys a green pencil' and some other 'Jones buys a yellow pencil'. If e.g.
the first one is true and the second one false in the ideal world, there seems
to be no moral reason why it should cease to be ideal if the truth values
were reversed. So norm systems with one-element bases may be looked
upon with some suspicion.
IX. A PARADOX
Some theorems of SDL have been called paradoxes. This means of course
that they seem counterintuitive, although they are derived from intuitively
AN ANALYSIS OF SOME DEONTIC LOGICS 131
Ross' paradox tried to say that too many acts were obligatory. Other
paradoxes arise because we do not seem to be able to express certain
obligations in SDL. Among them is the paradox of contrary-to-duty
obligations, discussed by Roderick M. Chisholm in [10]. I will closely
follow Lennart Aqvist's presentation of it in [7].
Consider the following four sentences:
I It ought to be that Smith refrains from robbing Jones.
II Smith robs Jones.
III If Smith robs Jones, he ought to be punished for robbery.
IV It ought to be that if Smith refrains from robbing Jones he
is not punished for robbery.
AN ANALYSIS OF SOMB DBONTIC LOGICS 133
DO o (tlf)
Dl - O(klf)
D2 OU n glh) +-+ OU Ih) & o (glh)
D4 0Ulh n g) -+ OU u - glh)
DS OU u glh) & - OU /h) -+ O(g/h n - f)
DS' If f u 9 ¥= t, then OU u g/h) & - OU/h)
-+ O(g/h n - f)
D6 OU/g) -+ OU/t)
D7 OU/k) -+ OU/g)
First of all we observe that DO-D2 say that if the circumstances are
constant, we have a subsystem isomorphic to SDL. For every J we can
find a basis for the subsystem where J describes the constant circum-
stances. Let the basis be B,. The other axioms then deal with the question
how the different bases are related to each other.
THEOREM 5: The following holds in Rescher's restricted system:
(i) ifJu 9 ¥=t and B" S;; Ju 9 and not B" s;; J, then B"nN' s;; g; (ii) Bt s;; B,
for every J; (iii) B, S;; Bk for every f.
Proof.' Immediate translations of DS', D6 and D7.
THEoREM 6: In the system DO-D2, DS' the following holds for every J:
either B, is t, or B, is a one-element set or there exist disjoint bases.
Proof.' Suppose B, is not t and that x and yare two distinct elements
in B,. By Lemma 1 there is a formula 9 such that B, s;; 9 c: t and by
Lemma 2 a formula h such that x belongs to hand y to - h. BI S;; (g n h)
u (g n- h)=g¥=t. B, is not a subset of 9 nh. By Theorem SCi) B,n-(,n")
is a subset of 9 n - h. The same argument with 9 n,.., h substituted for
gnh and vice versa proves that B,n-(,n-") is a subset of gnh, which
proves the existence of disjoint bases.
COROLLARY: In the system DO-D2, DS', D6 every basis is either t or
a one-element set.
Proof: By Theorem S(ii) Bt is a subset of every basis, which excludes
disjoint bases.
136 BENGT HANSSON
EI-E4 is the axiom system which von Wr!ght presented in [17]. In [18]
he changed EI to El'. The reason for this will be discussed below. EO is
the same innocuous axiom as AO, BO and DO:
EO o (t/I)
EI '" O(k/I)
El' '" O(k/t)
E2 OU n g/h) +-+ OU /h) & o (g/h)
E3 OU/g) & OU/h) -+ oU/g u h)
E4 OU /g u h) -+ OU /g) & OC! /h)
138 BENGT HANSSON
then there cannot be a duty to see to it that the window is open should
the sun be shining. This is manifestly absurd. Generally speaking: From
a duty to see to a certain thing under certain circumstances nothing can
follow logically concerning a duty or not-duty under entirely different,
logically unrelated, circumstances."
What von Wright wants is then the existence of (or at least the possi-
bility of the existence of) two different circumstances 9 and h and a
formula/which is obligatory under 9 and forbidden under h, i.e. BII shall
be a subset of/and Bh a subset of '" f, i.e. BII and Bh shall be disjoint!
He provides for this by changing E1 to E1'.
Does this change really exclude the kind of situations he described?
I do not think so. True, the theorem mentioned is no longer a theorem;
but there still are semantical difficulties which are not expressible in
formulas. Let us call a circumstance/ abnormal if o (kl/) is true; other-
wise it is normal. The crucial point in the proof of the unwanted theorem,
as von Wright sees it, was that 9 n h was necessarily normal. Now assume
that there is at least one abnormal circumstance and that we accept El'.
Then the set T of all complements of abnormal circumstances is a
consistent theory. For suppose that / and 9 are in T, i.e. that 0 (k/ '" /)
and 0 (k/ '" g) are true. By E3 0 (k/ '" / u '" g) is true, i.e. 0 (k/ '" (f n g»)
is true. Therefore/ng is in T. Now suppose that/is in T and that/is a
subset of g. 0 (k/ '" /) can as well be written in the form 0 (k/ '" / u( '" /
n,.., g») which by E4 implies O(k/ '" / n '" g) which is O(k/ '" g). There-
fore 9 is in T. k is not in T because t is normal by El'. Therefore T is
consistent. Let its non-empty basis be B and let x denote an arbitrary
but from now on fixed element in B. A cir~umstance J is normal exactly
when its complement is not in T, i.e. when B is not a subset of ,...., f, i.e.
when / has non-empty intersection with B. As a special case, every
formula containing x, forms a normal circumstance.
Now let / and 9 be any two formulas. One of the formulas / ng,
/ n '" g, '" / n g, '" / n,..., 9 is true in the valuation x. Therefore this
formula forms a normal circumstance. It is no longer true that every circum-
stance is normal, but there is at least one normal circumstance in· each qua-
druple of the type mentioned. If/and 9 are 'entirely different, logically
unrelated', it seems safe to say that the components of the three other for-
mulas are so too. So there is still a wealth of logically unrelated circum-
stances, such that whatever is obligatory under one of them is permitted
140 BENGT HANSSON
under the other. This is contrary to what von Wright saiti in the passage
quoted above. In general: if obligations under circumstance f are to be
logically independent of obligations under circumstance g, then f n 9 has
to be false in every valuation in B, i.e. f n 9 has to be abnormal. If von
Wright wants obligations under two circumstances to be independent as
soon as the circumstances are independent, then the intersection of two
independent circumstances must always be abnormal, which in turn
contradicts even El'.
The system EO, El', E2-E4 allows conflicting duties. This would be
a shortcoming of a logic of primary duties, but, von Wright argues, it is
a virtue in the case of a logic of conditional duties. Consider e.g. some-
one who has promised to do f, while f is forbidden. He is obligated to
refrain from f, but also to do f because of his promise. It may be that
this is a good thing taken by itself, but it is not when combined with E2,
because then everything becomes obligatory as soon as we have con-
flicting duties. And I do not think that von Wright intended to say that
someone is obligated to murder his uncle only because he promised to
punch his brother on his nose next time he caught sight of him.
But changing El to El'is not the only way to allow disjoint sets. One
can e.g. drop or change E4 instead. I think that this is a more attractive
way, since E4 seems to have some counterintuitive consequences. Let
the circumstances be that someone tries to save somebody from drown-
ing. He has succeeded in landing the man. If the man is unconscious he
is obligated to give him artificial respiration, but if the man is dead he is
not. Since only one of o (//g) and o (//h) is true, it cannot be true that
O(//g uh), i.e. if the man is dead or unconscious (and one does not know
which) there is no obligation to give him artificial respiration.
An example borrowed from Lawrence Powers' [12] also questions the
validity of E4. John Doe has got Suzy Mae pregnant. According to some
system of norms (admittedly not commonly accepted nowadays) he is
obligated to marry her under these circumstances. But, by theorem 10,
BIn, is a subset of B/ . Therefore everything obligatory under f is still
obligatory underf ng. So John Doe is still obligated to marry Suzy Mae
whatever may happen in addition to the circumstances already described.
In fact John shot poor Suzy through the head when he heard about her
condition, so he is now obligated to marry a dead girl.
Although von Wright's system admits more models than Rescher's,
AN ANALYSIS OF SOME DEONTIC LOGICS 141
it still has unpleasant consequences. I think that the system will improve
if disjoint bases are provided for in another way than changing El to El'.
How this can be done will be discussed in the next few sections.
in those logics. What does it mean to say O( '" Ilf)? Let/'be 'Smith robs
Jones'. It seems rather pointless to say 'Smith ought to refrain from
robbing Jones in the circumstance where he actually robs him'. If Smith
has robbed Jones, he cannot 'undo' it. He can restore what he robbed
(and this is obligatory under circumstance f in normal norm systems)
but this act is not the act of refraining from robbing Jones. We may
perhaps claim that the sentence in question only means that he should
not have done what he did, but then there would be no reason to mention
the circumstances; no matter what he actually did, he should not rob Jones.
Perhaps the situation is different iff is an act like 'Smith is smoking
in a no-smoking car'. We could then read D(......, flf) as 'if Smith is
smoking in a no-smoking car, he ought to stop'. This sounds good, but
there are some fine points here. Let us discuss this and other proposed
readings of 0 ( . . . , flf) systematically.
First, dyadic obligations are secondary, reparational obligations, telling
someone what he should do if he has violated (intentionally or not) a
primary obligation. Therefore they should not merely say that the agent
should not have done what he did; the primary obligation 0 ( ......, f)
already said that and the situation would be completely described by the
mixed formulaf &O( '" f) if one wants to stress that the agent actually
violated the 9bligation. If one takes conditional obligations seriously,
one has to realize that an agent cannot 'undo' what he has actually done.
The best way to read O(flg) is then 'now that the agent actually has
done g, he ought to do f'.
Second, if we return to the smoking Smith, it seems to make sense to
say 'now that Smith actually has smoked in a no-smoking car, he ought
to refrain from smoking in a no-smoking car', because it does not
necessarily mean that he shall 'undo' the smoking already done, but only
that he shall refrain from further smoking. More carefully stated the
obligation then reads: 'now that Smith has smoked in a no-smoking car
up to this moment he ought to refrain from continuing after this moment'.
But this is not an obligation of the form 0(......, Ilf).
From this I conclude that formulas like O(I/g) shall never be true if
f and g are disjoint, if circumstances are taken seriously. And by this I
mean that the circumstances are regarded as something which has actually
happened (or will unavoidably happen) and which cannot be changed
afterwards.
AN ANALYSIS OF SOME DEONTIC LOGICS 143
(5) O(j (\ gjh) +-+ O(j jh) & o (gjh)( = D2, E2).
Valid in all three systems. This formula reflects the idea of a basis and
since the DSDL systems are constructed as dyadic extensions of SDL it
should be valid.
(7) O(j u gjh) & '" o (jjh) -+ O(gjh (\ '" f)(= DS).
Valid in DSDL3 but not in DSDLl or DSDL2. This is the unrestricted
version of Rescher's fifth axiom. Anderson's problem in this connexion
is solved by the considerations in Section XIII.
University of Lund
AN ANALYSIS OF SOME DEONTIC LOGICS 147
BIBLIOGRAPHY
NOTE
• Reprinted by permission of the author and the publisher from Nous 3 (1969) 373-398.
KRISTER SEGERBERG
The topic of this paper is the logical relationship between the notions of
commitment and obligation. When this question was first raised it was
thought that the former notion could be analysed in terms of the latter
and the Boolean connectives of classical logic. However, in the end
almost every effort to carry out such an analysis has led to intuitively
unacceptable consequences, and there would seem to be substantial
agreement among today's logicians that the notion of commitment
cannot be reduced to that of obligation. 1 In this paper we define a family
of propositional logics in which both commitment and obligation are
formalized as independent operators.
The words 'commit' and 'commitment' have several uses in English
only some of which are of interest to deontic logic. Even among the latter
there is an obvious distinction to make. In one sense commitment is a
one-place concept; in this sense, we believe, 'commitment' is to be
identified with 'obligation'. In another sense - its 'proper' sense - it is a
two-place concept which is rendered by '(the fact) that A commits the
agent to (the proposition that) B'. (Actually, it may be argued that both
concepts of commitment ought to be relativized to agents in order that
their full logical form be brought out. According to this view commitment
in the first sense is an at least two-place concept, and 'proper' commitment
is at least four-place: 'By virtue of (the fact that) A, g commits b to the
proposition that B'. In this paper, though, neither g nor b will play a
significant role. More precisely, g will be entirely neglected and b will be
suppressed; unless we specify otherwise it may be assumed that b is an
arbitrary fixed agent.)
We shall use A com B for the proposition that A commits the agent
to B. Another way of expressing the same proposition would be to say
that if A then the agent is committed to B, but then it must be noted that
'committed' is used in the sense of 'obligated' and that the 'if-then'
cannot be identified with material implication.
R. Hilpi""" (ed.). Deo"tic Logic: Introductory and Sy.tellJfJtic Relllllng•• 148-158. All right. reserved.
CopyrlghtO 1970 by D. Reidel Publishl"g Comptmy. Dordrecht-Holland.
SOME LOGICS OF COMMITMENT AND OBLIGATION 149
II
Throughout the paper we shall use the abbreviation IICII'- for the exten-
sion ofC in tW, that is, for the set {U:UE U &F~C}. Note thatthe clause SO
is equivalent to the condition that A com B if and only if there is some
neighborhood IX of u such that IX £; IIA -+ OBII'-.
We say that A is true in a model if F"A for every u in the domain of the
model. The set of formulas true in all models can be axiomatized as
follows. Let our rules of inference be the following three:
A A-+B
R1 (Modus Ponens)
B
A
R2 (Obligation)
OA
(A -+ OB) -+ (A' -+ OB')
R3°
A com B -+ A' com B'
Our axioms are the tautologies of classical propositional logic plus all
instances of the familiar schema
Let this basic calculus be called BO. It can be shown, for example with
the now so common Henkin type of argument, originally used by
Makinson and Scott, that a formula is demonstrable in BO if and only
if it is true in every model.
From the point of view of our intended interpretation of 0 as obli-
gation and com as commitment this calculus seems too weak. It is stan-
dard among deontic logicians to demand that falsehood not be obligatory.
Moreover, it seems to this author that one cannot help being committed
to what is logically true, although no doubt some philosophers would
disagree. Finally, one of the basic facts about the kind of commitment
we have in mind is that if A commits one to B and if A is the case, then
SOME LOGICS OF COMMITMENT AND OBLIGATION 153
*1 0.1 -+ .1
*2 AcomT
*3 A com B -+ (A -+ OB).
Let F O be the calculus obtained by adding all instances of :#= 1-3 to *
the stock of axioms of BO, keeping the three rules as before. We regard
FO as a fundamental deontic logic of obligation and commitment. It is
not difficult to show that a formula is demonstrable in pO if and only if
it is true in all models that satisfy these conditions:
The point is that com may be defined in terms of c:J and the other primi-
tives:
A comB = df c:J (A ~ OB)
(Note how well this fits with reading 'A com B' as 'if A then the agent is
committed to B'!) With the new set of primitives we obtain a more
elegant axiomatization of the set of formulas true in all models. Let B
be the basic calculus whose axioms are the tautologies of classical logic
*
and the instances of schema 0, and whose rules are Modus Ponens,
Obligation, and the new rule
A~B
R3
One can prove that the set of formulas true in every model is exactly the
set of formulas demonstrable in B.
The new language is richer than the old one, for whereas com can be
defined in the new language, 0 cannot be defined in the old. However,
unlike com, the operator 0 does not seem to have an obvious counter-
part in a natural language like English. In view of our intuitive remarks
in the introduction one would like to read [JA as 'it holds deontically
essentially that A' or something of this sort. But for such a reading of 0
the calculus B is not strong envugh; at least we must add as axioms to B
*
all instances of 1 and of
* *
by adjoining all instances of 5 and 6 to FK if and ('nly if it is true
in every Kripke model satisfying condition (i), the condition that S is
reflexive and the additional conditions
(v) V u V v V w(ifuSv and vRw then uRw).
(vi) V u V v V w(if uSv and uRw then vRw).
The schemata
*.7 'O'OA-A.
*8 OA-OOA
are well known; their 'corresponding' conditions a,re
(vii) S is symmetric.
(viii) S is transitive.
The set of formulas demonstrable in the calculus obtained by adding all
instances of ** 5-8 to FK is of course identical with the set of formulas
true in every Kripke model in which (i), (v), (vi) are satisfied and S is an
equivalence relation. Remembering that the set of S-alternatives of a
point may be thought of as a neighborhood of that point - in fact the
smallest one -.we observe that those models have the following interesting
property: for every point u there is a neighborhood such that throughout
the neighborhood exactly the same obligations and the same commit-
ments hold as in u. Thus such a model is partitioned into equivalence
classes within each of which non-deontic facts may vary but not deontic
ones. If one accepts the underlying logic one may perhaps say that this
concept of model explicates the concept of a deontic situation. 3
III
Ek promises to lend him his for the day we may say that Ek is committed
to giving Nilsson the car when he comes to pick it up, but few would say
that such a commitment obtains if it turns out that Nilsson shows
up drunk. It must be admitted that none of our logics can handle such
cases.
The most natural reply is, perhaps, that at least one of the com-operators
in Aqvist's example is of a kind which our semantics does not formalize.
Let the exacting notion of commitment we have analysed be called 'strong
commitment'. Aqvist's example, it seems, involves a kind of 'commit-
ment-reasonable-in-principle', or 'weak commitment' for short. That
A 1\ B strongly commits me to C if A does, seems acceptable - as accept-
able as that there are cases when A, but not A 1\ B, weakly commits me
to C. If so, Aqvist's interesting observation ceases to be an objection
against our theory.
It would be tempting to identify strong and weak commitment with
absolute commitment and prima facie commitment, respectively.4 The
development of semantics for weak or prima facie notions of commitment
within the general framework defined above seems like an interesting
problem, but in this paper it is left open.
Abo Academy
BIBLIOGRAPHY
[1] Alan Ross Anderson, 'On the Logic of "Commitment"', Philosophical Studies 10
(1959) 23-27.
[2] Lennart Aqvist, 'A Note on Commitment', Philosophical Studies 14 (1963) 22-25.
[3] Brian F. Chellas, The Logical Form of Imperatives, Perry Lane Press, Stanford,
Calif., 1969.
[4] Risto Hilpinen, 'An Analysis of Relativised Modalities', in Philosophical Logic (ed.
by J. W. Davis et al.), D. Reidel, Dordrecht, 1969, pp. 181-193.
[5] Jaakko Hintikka, Models for Modalities. Selected Essays, D. Reidel, Dordrecht,
1969.
[6] David Lewis, 'Semantic Analysis of Subjunctive Conditionals', Xerox copied,
V.C.L.A., 1968.
[7] Richard Montague, 'Pragmatics', in Contemporary Philosophy. A Survey. I. Logic
and Foundations of Mathematics (ed. by R. Klibansky), La Nuova Italia Editrice,
Firenze, 1968, pp. 102-122.
[8] Dana Scott, 'A Logic of Commands', mimeographed, Stanford University,
Stanford, 1967.
[9] Dana Scott, 'Advice on Modal Logic', in Philosophical Problems in Logic: Some
Recent Developments (ed. by Karel Lambert), D. Reidel, Dordrecht, 1970, pp. 143-
173.
158 KRISTER SEGERBERG
NOTES
Deontic logic was, in origin, an off-shoot of modal logic. It got its decisive
impetus from observations of some obvious analogies between the modal
notions of necessity, possibility, and impossibility on the one hand and
the deontic or normative ideas of obligation ('ought to'), permission
('may'), and prohibition ('must itot') on the other hand. In a broad sense,
both groups of concepts can be called modal; the members of the first
group are sometimes referred to as alethic, those of the second group as
deontic modalities. (A preferable alternative to the term 'alethic' is
perhaps the term 'anankastic'.)
Beside analogies and similarities, however, there are also a number of
striking dissimilarities between the two types of modalities. Many of the
problems which have beset deontic logic since its birth are related to
these discrepancies. One difference is the absence in deontic logic of an
analogue to the principle 'Np -+ p' of modallogic.1 That which necessarily
is the case is also as a matter of fact the case; but that which ought to be
the case is far from being always actually the case. Another formal
difference between modal and deontic logic is that, whereas it is obvious
that the tautology necessarily is true ('Nt'), it is not intuitively clear that
the tautology also ought to be true ('Ot'). The idea expressed by COt' does
not seem to make good sense. By contrast, ',.., 0 ,.., t' seems not only
to make sense, but also to be true. This formula says that a contra-
dictory state of affairs is not a state which ought to be the case. This
is an analogue to the principle ',.., N,.., t' of modal logic which can
also be written 'Mt'. The principle '", N,.., t' is a weaker form of the
principle 'Np -+ p' in as much as the first follows logically from the
second (and principles of ordinary propositional logic, PL), but not
vice versa.
One can display these analogies, and failures of analogy, in the follow-
ing table:
R. Hilpinen (ed.), Deont/c Logic: Introductory and SYltematlc Reodlng8, 159-177. A.1l right8 re8er.ed.
Copyright C 1970 by D. Reltkl Publl.lhlng Company, Dort/recht·Holland.
160 GEORG HENRIK VON WRIGHT
A2
A3 Nt Not: Ot
II
and the one which excludes 'Ot' from the status of a theorem - are em-
bodied in the following modified definition:
'Nc(p, q)' = de 'N(q ~ p) & M,.., p & Mq'.
Let the symbol 'C' signify contingency. If we accept the modified
definition of a necessary condition, we can prove the theorem 'Nc(p, q)
~ Cp&Cq'. From our definition of 'ought' we then also prove 'Op~ Cp'
and, since contingency entails possibility, 'Op~Mp'. The last may be
regarded as a version of the principle, commonly associated with the
name of Kant, that 'Ought implies (entails) Can'.
I shall call the addition 'M,.., p&Mq' to the definition of Nc a contin-
gency-clause.
Someone may wish to object to the new definition that it is an ad hoc
modification for the sake solely of accommodating deontic logic within
a theory of conditions. My rejoinder will consist of two parts:
First, I do not think the objection a fair one. Quite apart from con-
siderations relating to deontic logic and normative concepts, the suggested
modification seems to me reasonable. But it challenges the serious
problem of how to deal with relations of conditionship between non-
contingent, i.e. necessary or impossible, propositions. Since this problem
is not relevant to deontic logic, we need not discuss it at length here. Let
it only be said that I think its solution has to be given in terms of higher
order (iterated) modalities. A necessary condition of a necessary propo-
sition, I submit, is not a necessary condition of the truth of that propo-
sition, but of its necessity; and a necessary condition of an impossible
proposition is not a necessary condition of the falsehood of that propo-
sition, but of its impossibility.
Secondly, instead of modifying the definition of a necessary condition
we could correspondingly modify the definition of the O-operator by
adding to it the requirement that the states 'p' and 'f' should be contin-
gent. This would seem entirely unobjectionable and would lead to essen-
tially the same results as far as deontic logic is concerned. But, as indi-
cated, I should favour the more daring modification proposed above.
Let it be observed in passing that 'Op~ O(p v q)' is not a theorem of
the deontic logic we are building. This is the famous Ross's Paradox
formula. The fact that it is not a theorem does not mean, however, that
the notorious troubles caused by this paradox, and its variations, have
164 GEORG HENRIK VON WRIGHT
Both under the unmodified and the modified versions of the definitions
of 'Ne' and 'Se' these equivalences hold true: 'Se(p, q)+-+Ne(q,p)' and
'Seep, q)-Se(,.., q, ,.., p)' and 'Ne(p, q)-Ne(,.., q, ,.., p)'.
Accepting the modified definition of 'Se', we have to note a restriction
on the validity of the P-distribution principle, analogous to the restriction
we noted in Section III for the O-distribution principle. It does not follow
logically from the fact that the disjunction of two states is contingent that
each one of the states individually is contingent; one of them may be
impossible. The implication 'pep v q)~Pp&Pq' therefore requires a
conditional clause 'Mp&Mq'. On the other hand, if both of two contin-
gent propositions strictly imply a third contingent proposition ('I'), then
their disjunction will necessarily be contingent. The implication 'Pp &Pq
~P(p v q)', therefore holds unconditionally.
Assume that 'p' is something which, in the strong sense, may be and
that 'q' is something which ought to be. Then we have, according to our
definitions, 'N(p~I)&N(I~q)'. This entails 'N(p .... q)' which entails
'p~q'. Thus if something which, in the strong sense, may be the case
actually is the case, then everything which c!lght to be the case is the case,
too. Of that which is in the strong sense permitted one can, in other
words, avail oneself only on condition that none of one's obligations is
thereby violated. This is a refiexion, in deontic terms, of the general
principle for conditions which says that a sufficient condition of some-
thing can be realized only provided that all the necessary requirements
for the occurrence of the thing in question are satisfied as well. To say
of something that it is in the strong sense permitted without including
in the description mention of all dutybound things is therefore an elliptic
mode of speech. Similarly, to say that something is a sufficient condition
of something else without mention of the necessary conditions is elliptic,
too. This elliptic mode of expression is, however, quite commonly used.
166 GEORG HENRIK VON WRIGHT
sufficient conditions obtains too. The idea that, for a given state 'p', there
exists some set S in which 'p' is determined captures, I think, an impor-
tant aspect of the notion of determinism. Similarly, the idea that, for all
contingent states, there is some such set captures an aspect of the idea
of Universal Determinism. s
Consider next a normative code or order or system which stipulates
that certain contingent states, 'p', 'q', ... , are necessary conditions of a
contingent state 'I' which is not itself one of those contingent states, nor
a truth-functional compound of some of them. Assume that the conjunc-
tion of all those necessary conditions of 'f' is a sufficient condition of '1'.
'I' is determined, in other words, in the set of obligations stipulated by
the code (in terms of 'I'). This assumption means that, if everything that
according to this code ought to be the case actually is the case, then
whatever else there is (happens to be) may be the case. Or, speaking in
juristic terms: if all obligations are satisfied, then whatever else is done
(or omitted) is also allowed. Then we shall say that the normative code
or order or system is closed with regard to the obligations defined in the
terms of 'J'. This idea of the closedness of a legal (or other normative)
order is thus a special case of a more general idea of determinism.
An example will be given to illustrate the above. Assume that 'I' is
determined in some S and that the conjunction of all its necessary con-
ditions in S is equivalent with 'p&q'. Then we have 'O(p&q)' and
'P(p&q)'. Consider now a state Or' which is, as I shall say, co-contingent
with 'p&q'. By this I mean that 'M(p&q&r)' and 'M(p&q&- r)' are
both ofthem true. Then we can derive 'P(p&q&r)' and 'P(p&q&- r)'.
Of course, neither or' nor its contradictory' - r' is by itself a sufficient
condition of 'J'. cPr' and 'P"" r' are not derivable. But we can introduce
here a new piece of terminology and say that, if a state of affairs is co-
contingent with the conjunction of all the necessary conditions of a
determined state 'I', then this state and also its contradictory state are,
in the frame or setting of the necessary conditions, sufficient conditions
of 'J' and in this sense also strongly permitted. (Cf. what was said in
Section IV on the elliptic use of 'permitted' and 'sufficient condition'.)
Since, moreover, neither the state in question nor its contradictory is
itself a necessary condition of '1' - this would conflict with the assumption
of co-contingency - both states are in the weak sense permitted. '"" 0 "" r'
and '- Or' are derivable.
168 GEORG HENRIK VON WRIGHT
VI
the state 'r'. If we make the assumption that it is possible for both the
two states, 'p' and 'q', to be absent, i.e. if we assume that their disjunction
is not in itself a necessity, then the above second order expression entails
that the disjunction of the state 'p' and the state 'q' is a necessary con-
dition of the state 'r' - though not viee versa. We have, in other words,
'Se( - p, Ne(q, r»)&- N(p v q)~Ne(p v q, r)'.
Now substitute 'J' for Or'. Then we get from the last formula
'See - p, Oq)&- N(p v q)~ O(p v q)'. The antecedent, I suggest, is an
expression for commitment (relative, conditional obligation). I shall
define a new deontic operator Q as follows:
'Q(qjp)' = df'Se(p, Oq) & - N(p ~ q)'.
'Q(qJp)' can be read 'it ought to be the case that q, given that p'.
Assuming that it is in the agent's power to produce 'p', then by producing
this he 'becomes' obligated to produce also 'q', unless 'q' is something
which is of necessity there as soon as 'p' is there. (For in this last case,
it 'makes no sense' to speak of an 'obligation'.)
'Q(qjp)' entails 'O(p~q)' but not viee versa. 'O-p' does not entail
'Q(qjp)'. Nor does 'Oq' do this.
Of particular interest is the case when '0 - p' actually obtains with
'Q(qjp)'. Then the second expression tells us what the agent ought to do,
when he has done something he was, in fact, forbidden to do. This sort
of commitment or conditional obligation is of the kind for which Pro-
fessor Chisholm has coined the name Contrary-to-Duty Imperative. 6
The above ideas seem to me to lead to a satisfactory account of the
notion of commitment, and of conditional norms generally. Some
problems of old standing in deontic logic now acquire a natural solution. 7
VII
order condition concept. In it is involved the idea that the fact that
something is a necessary condition of something else is itself a necessary
condition of something.
If we let these things of which something is a necessary condition be
identical with the state of affairs 'J' which figures in our definitions of the
deontic operators, we obtain higher order deontic expressions such as
'OOp', 'OPp', 'POp', 'PPp'. Thus, for example, 'OOp' when explicated
in terms of conditions means 'Nc(Nc{p, I), I)'. When further explicated
in modal terms we get from this the expression 'N(I --.. N{I --.. p) &MI &M
"'p)&MI&M", (N(I--"p) &MI&M "'p)'. If we ignore, or take for
granted the contingency-conditions attached to the antecedents and
consequents of the two strict implications involved, we can simplify the
formula to 'N(I--..N{I-+p)'.
In the light of this interpretation of higher order deontic expressions
one can profitably examine some formulae, whose status as logical truths
about the deontic modalities has been defended by some and disputed
by others. I shall consider one example. This is a formula which I shall
label Prior's Formula. 8
It is clear that 'Op --.. p' cannot be a truth of logic. (Cf. Section I.) But
what about 'O{Op-+p)'? In words this formula says that it ought to be
the case that that which ought to be the case actually also is the case. This
sounds reasonable enough. Let us see what it amounts to when the
formula is translated into the terminology of conditions and modalities.
As a statement of conditionship the formula means 'Nc({Nc(p, I)
--.. p), I)'. If we write it out in modal terms we get 'N(I --.. (N(I --.. p) &MI
&M,.., p--..p) &MI&M(N{I--..p)&MI&M ""p&"" p)'.
Let us first ignore the contingency-clauses and 'pull out' the strict im-
plication part of the formula. Then we get 'N(I --.. (N(I --.. p) --.. p)'. This
is easily shown to be a theorem of modal logic (M). Therefore the condi-
tional statement which says that, if the contingency-clauses are satisfied,
then Prior's formula holds good, is a theorem of modal logic, too.
We now tum attention to the contingency-clauses. The third member
of the main conjunction in the modal expansion of Prior's formula can
be abbreviated to 'M ( Op &,.., p)'. It says that it is possible that the state
of affairs 'p' ought to be but nevertheless is not.
But is this clause needed at all? Is it not trivially satisfied by virtue of
the fact that 'Op --.. p' is not a theorem? The answer to the last question
DEONTIC LOGIC AND THE THEORY OF CONDITIONS 171
is negative. The mere fact that 'Op-+p' and therewith also 'N(Op-+p)'
is not a theorem does not make the negation of 'N(Op-+p)' which is
equivalent to 'M(Op&- p)' a truth of modal logic. Whether the clause
upon which Prior's formula is conditional is satisfied or not is therefore
a matter of contingent truth and not of logical necessity.
What the 'if-clause' which is needed to warrant the truth of Prior's
formula does, interestingly, is to draw attention to the fact that, although
it is inherent in the notion of an obligation that the obligatory is some-
thing in itself contingent and in this sense neglectable, it may yet as
obligatory be impossible to neglect. cOp -+ M - p' is a theorem of deontic
(modal) logic. But 'Op&M(Op&- p)' can be true or false, and so can
'Op&N(Op-+p)'. When 'Op&M(Op&-p)' is true ofa state 'p', J shall
say that the fact that 'p' ought to be is a neglectable obligation. When
again cOp & N (Op -+ p)' is true of a state I shall say that the fact that this
state ought to be is a non-neglectable obligation.
I conclude this section with the conjecture and sugcstion that the
distinction which we have discovered between two types of obligation
can be interestingly related to things familiar from traditional ethical
theory. (Over-riding obligations, primajacie obligations, etc.)
VIII
the weak 'may'. Similarly in the case when I say that in order to get to a
destination in time I need not take a tram, but may also use some other
means of transportation. Assume, however, I said that in order to get
there in time, I may take a bus. This would normally mean that, if I go
there by bus I shall arrive in time, but that there perhaps are other means
beside this to secure the same end of my action. I may, e.g., also take
the subway, I can choose between it and the bus. This is the strong 'may'.
The problem of the content of 'j' is of particular interest to philosophy,
when the 'ought' (and the two 'may') is a legal or a moNI 'ought' ('may').
I would submit for consideration that for an important type of lepl
'ought' - the 'ought of lepl obligation' as I shall call it - 'j' is a state of
affairs which can be characterized as immunity to pIIfIislunmt (a punitive
reaction on the part of a legal machinery). The actions which it is our
legal duty or obligation to do are those which are required of us if we
are to be immune to lopl punishment. They are necessary to ensure
immunity, but whether they are also sufficient depends upon whether the
legal order, or part of legal order, under consideration is closed or open.
(Cf. above Section Y.)
But what is immunity? This is itself a problematic notion and can be
understood in several ways. One could suggest this: Immunity means
that, unless 'p' is (done), punishment will follow. But this would hardly
be an interesting notion of immunity. For 'p' may be neglected and yet
no punishment follow, either because the criminal is not caught or,
although caught, acquitted because not proved guilty. To be immune to
punishment for neglect of 'p' means rather that one cannot be legally
punished for having neglcctecl 'p'. And the contradictory of immunity,
which can also be called liability to punishmeDt, means that one can be
thus punished. But what do 'can' and 'cannot' mean here? They are
modal notions and on that gt'ound alone related to 'ought' and'may'.
In the analysis of the notions of immunity and liability to punishment,
deontic notions may thus crop up anew. This would not entail circularity
or show that immunity and liability cannot be interestingly used for
defining the notions of legal obligation and permission. But it would
show that in the structure of a legal order other types of 'ought' and
'may' than (legal) obligations and permissions· are involved - and it
would chal1enge further analysis of how the various types of norm which
build up this order are related and intertwined.
174 GEORG HENRIK VON WRIGHT
IX
trying to settle it until we have first stated/or what the 'this' and the 'that'
are thought to be required. If they are required for different things, there
can be no dispute between the two 'oughts' (but perhaps another dispute
about which ofthese two things ought to be pursued). If they are required
for the same thing (the same '1'), one of the disputants may be right and
the other wrong or both may be wrong. And now 'right' and 'wrong' is
a matter of truth and falsehood.
Even when the elliptic character of a given 'ought' (or 'may') is recog-
nized in principle it may be difficult or even impossible to specify in
practice the '1' relative to which something ought to or may be. Perhaps
we only have some dim conception of its nature. (We had been taught to
think the matter was clear without questioning. This is what all too often
happens in an authoritarian type of society.) Then we may not come to
a 'grip' with the question of truth in connexion with the 'ought' ('may')
and the norm takes on an 'alogical' appearance.
A characteristic use of 'ought' (not so much it would seem of 'may')
is to evince an evaluation. Thus in saying "a good fountain-pen ought
not to leak" or simply "a fountain-pen ought not to leak", I may be
evincing a standard of goodness in fountain-pens. "A leaking fountain-
pen, according to my conception ofthe matter, simply is not a good one."
To adopt this standard is to make the property of not leaking a necessary
requirement of goodness in fountain-pens.
Another characteristic use of 'ought' (and to some extent also of the
strong 'may') is for enforcing patterns of behaviour (conduct). This is,
in a paramount sense, the normative use of the words 'ought' and 'may'.
Imperatives (sentences in the imperative mood) are also used for the
same purpose. It would be a confusion to say that 'ought'-statements are
imperatives. Imperatives are not statements and imperatives, as has so
often been pointed out, are not true or false. But 'ought'-sentences are
commonly and characteristically used as imperatives, viz. used for the
purpose of urging (making) people to behave in a certain way.
It is futile to dispute whether 'open the window' and 'you ought to
open the window' mean (as it were 'intrinsically mean') the same or not.
Sometimes someone addresses another person with the words "you ought
to open the window", when he could just as well have said "open the
window". But perhaps he will, when he says the former, more often than
not have 'at the back of his mind' an idea of the thing for which com-
176 GEORG HENRIK VON WRIGHT
Academy of Finland
NOTES
• An earlier version of this paper appeared in Critica 2 (1968) 3-25. The author wishes
to acknowledge his indebtedness to Professor Carlos Alchourr6n (Buenos Aires) for
criticisms and suggestions which have been useful in revising the essay.
1 I shall not explain here my use of symbols nor the conventions adopted about
brackets. I shall assume that the symbolism will be either familiar to the reader or else
self-explanatory .
2 I shall use the words 'obligatory', 'obligatoriness', 'permitted', and 'permittedness'
with a loose and wide meaning which is roughly equivalent with ordinary uses of
'ought' and 'may'. There is also a stricter use of the words mentioned and of the terms
'obligation' and 'permission', which belongs typically in legal and moral contexts.
3 'A Reduction of Deontic Logic to Alethic Modal Logic', Mind 67 (1958) 100-103.
4 Another treatment of the elements of a logic of conditions is found in my book A
Treatise on Induction and Probability, Routledge and Kegan Paul, London, 1951. That
treatment is framed, not in modal propositional logic, but in the monadic lower
functional calculus (theory of properties and quantifiers).
5 Cf. A Treatise on Induction and Probability, Chapter Ill, Section 2.
8 R. M. Chisholm, 'Contrary-to-duty Imperatives and Deontic Logic', Analysis 24
(1963) 33-36.
7 Cf. also my paper 'A New System of Deontic Logic' (in this volume, pp. 105-120.
It is readily shown that, subject to the appropriate contingency conditions, the dyadic
expressions 'Q(f)' satisfy the Axioms BI-B3 given in the original version of the paper.
The undesirable consequence which led me to abandon Bl in favour of Bl' does not
arise, if the notion of a conditional obligation is explicated, as here, in the terms of
conditionship relations between contingent propositions. The problem of predica-
ments or conflicting duties will now have to be treated differently, viz. along the lines
suggested in my paper 'Deontic Logic' (Mind 60 (1951) 1-15). If a man, like Jephtha,
commits himself to do something which is forbidden, then the committing action was
itself forbidden. 'Q(q/p)' in conjunction with '0,.., q' entails '0 ,.., p'.
8 It was introduced into the literature by A. N. Prior in his Formal Logic, Clarendon
Press, Oxford, 1955.
INDEX OF NAMES